EP1449108A4 - Classement de caracteristiques pretraitees pour une machine a vecteur de support - Google Patents

Classement de caracteristiques pretraitees pour une machine a vecteur de support

Info

Publication number
EP1449108A4
EP1449108A4 EP02778747A EP02778747A EP1449108A4 EP 1449108 A4 EP1449108 A4 EP 1449108A4 EP 02778747 A EP02778747 A EP 02778747A EP 02778747 A EP02778747 A EP 02778747A EP 1449108 A4 EP1449108 A4 EP 1449108A4
Authority
EP
European Patent Office
Prior art keywords
data
features
genes
svm
training
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Withdrawn
Application number
EP02778747A
Other languages
German (de)
English (en)
Other versions
EP1449108A1 (fr
Inventor
Jason Weston
Andre Ellisseef
Bernhard Schoelkopf
Fernando Perez-Cruz
Isabelle Guyon
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Health Discovery Corp
Original Assignee
Health Discovery Corp
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Health Discovery Corp filed Critical Health Discovery Corp
Publication of EP1449108A1 publication Critical patent/EP1449108A1/fr
Publication of EP1449108A4 publication Critical patent/EP1449108A4/fr
Withdrawn legal-status Critical Current

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N20/00—Machine learning
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
    • G06F18/211—Selection of the most significant subset of features
    • G06F18/2113—Selection of the most significant subset of features by ranking or filtering the set of features, e.g. using a measure of variance or of feature cross-correlation
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
    • G06F18/2411—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on the proximity to a decision surface, e.g. support vector machines
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N20/00—Machine learning
    • G06N20/10—Machine learning using kernel methods, e.g. support vector machines [SVM]
    • G—PHYSICS
    • G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
    • G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
    • G16H50/00—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics
    • G16H50/20—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics for computer-aided diagnosis, e.g. based on medical expert systems

Definitions

  • the present invention relates to the use of learning machines to identify relevant patterns in datasets containing large quantities of diverse data, and more particularly to a method and system for selection of features within the data sets which best enable identification of relevant patterns.
  • human Genome Project has completed sequencing of the human genome.
  • the complete sequence contains a staggering amount of data, with approximately 31,500 genes in the whole genome.
  • the amount of data relevant to the genome must then be multiplied when considering comparative and other analyses that are needed in order to make use of the sequence data.
  • human chromosome 20 alone comprises nearly 60 million base pairs.
  • Several disease-causing genes have been mapped to chromosome 20 including various autoimmune diseases, certain neurological diseases, type 2 diabetes, several forms of cancer, and more, such that considerable information can be associated with this sequence alone.
  • Proteomics is the study of the group of proteins encoded and regulated by a genome. Genomic activation or expression does not always mean direct changes in protein production levels or activity. Alternative processing of mRNA or post- transcriptional or post-translational regulatory mechanisms may cause the activity of one gene to result in multiple proteins, all of which are slightly different with different migration patterns and biological activities.
  • the human proteome is believed to be 50 to 100 times larger than the human genome. Currently, there are no methods, systems or devices for adequately analyzing the data generated by such biological investigations into the genome and proteome.
  • Machine-learning approaches for data analysis have been widely explored for recognizing patterns which, in turn, allow extraction of significant information contained within a large data set that may also include data consists of nothing more than irrelevant detail.
  • Learning machines comprise algorithms that may be trained to generalize using data with known outcomes. Trained learning machine algorithms may then be applied to predict the outcome in cases of unknown outcome, i.e., to classify the data according to learned patterns.
  • Machine-learning approaches which include neural networks, hidden Markov models, belief networks and kernel-based classifiers such as support vector machines, are ideally suited for domains characterized by the existence of large amounts of data, noisy patterns and the absence of general theories. Support vector machines are disclosed in U.S. Patents Nos. 6,128,608 and 6,157,921, both of which are assigned to the assignee of the present application and are incorporated herein by reference.
  • feature selection The quantities introduced to describe the data that is input into a learning machine are typically referred to as “features”, while the original quantities are sometimes referred to as “attributes”.
  • One method of feature reduction is projecting on the first few principal directions of the data. Using this method, new features are obtained that are linear combinations of the original features.
  • One disadvantage of projection methods is that none of the original input features can be discarded.
  • Preferred methods incorporate pruning techniques that eliminate some of the original input features while retaining a minimum subset of features that yield better classification performance. For design of diagnostic tests, it is of practical importance to be able to select a small subset of genes for cost effectiveness and to permit the relevance of the genes selected to be verified more easily.
  • the present invention comprises preprocessing a training data set in order to allow the most advantageous application of the learning machine.
  • Each training data point comprises a vector having one or more coordinates.
  • Pre-processing the training data set may comprise identifying missing or erroneous data points and taking appropriate steps to correct the flawed data or as appropriate remove the observation or the entire field from the scope of the problem.
  • preprocessing includes reducing the quantity of features to be processed using feature selection methods selected from the group consisting of recursive feature elimination (RFE), minimizing the number of non-zero parameters of the system (£ 0 -norm minimization), evaluation of cost function to identify a subset of features that are compatible with constraints imposed by the learning set, unbalanced correlation score and transductive feature selection.
  • RFE recursive feature elimination
  • the features remaining after feature selection are then used to train a learning machine for purposes of pattern classification, regression, clustering and/or novelty detection.
  • the learning machine is a kernel-based classifier.
  • the learning machine comprises a plurality of support vector machines. A test data set is pre-processed in the same manner as was the training data set.
  • the trained learning machine is tested using the pre-processed test data set.
  • a test output of the trained learning machine may be post-processing to determine if the test output is an optimal solution based on known outcome of the test data set.
  • the present invention also provides for the selection of at least one kernel prior to training the support vector machine. The selection of a kernel may be based on prior knowledge of the specific problem being addressed or analysis of the properties of any available data to be used with the learning machine and is typically dependant on the nature of the knowledge to be discovered from the data.
  • Kernels are usually defined for patterns that can be represented as a vector of real numbers. For example, linear kernels, radial basis function kernels and polynomial kernels all measure the similarity of a pair of real vectors. Such kernels are appropriate when the patterns are best represented as a sequence of real numbers.
  • An iterative process comparing postprocessed training outputs or test outputs can be applied to make a determination as to which kernel configuration provides the optimal solution. If the test output is not the optimal solution, the selection of the kernel may be adjusted and the support vector machine may be retrained and retested. Once it is determined that the optimal solution has been identified, a live data set may be collected and pre-processed in the same manner as was the training data set to select the features that best represent the data. The pre-processed live data set is input into the learning machine for processing. The live output of the learning machine may then be post-processed by interpreting the live output into a computationally derived alphanumeric classifier or other form suitable to further utilization of the SVM derived answer.
  • a system enhancing knowledge discovered from data using a support vector machine.
  • the exemplary system comprises a storage device for storing a training data set and a test data set, and a processor for executing a support vector machine.
  • the processor is also operable for collecting the training data set from the database, pre-processing the training data set, training the support vector machine using the pre-processed training data set, collecting the test data set from the database, pre-processing the test data set in the same manner as was the training data set, testing the trained support vector machine using the pre-processed test data set, and in response to receiving the test output of the trained support vector machine, post-processing the test output to determine if the test output is an optimal solution.
  • the exemplary system may also comprise a communications device for receiving the test data set and the training data set from a remote source.
  • the processor may be operable to store the training data set in the storage device prior pre-processing of the training data set and to store the test data set in the storage device prior preprocessing of the test data set.
  • the exemplary system may also comprise a display device for displaying the post-processed test data.
  • the processor of the exemplary system may further be operable for performing each additional function described above.
  • the communications device may be further operable to send a computationally derived alphanumeric classifier or other SVM-based raw or post-processed output data to a remote source.
  • a system and method are provided for enhancing knowledge discovery from data using multiple learning machines in general and multiple support vector machines in particular.
  • Training data for a learning machine is pre-processed.
  • Multiple support vector machines, each comprising distinct kernels are trained with the pre-processed training data and are tested with test data that is pre-processed in the same manner.
  • the test outputs from multiple support vector machines are compared in order to determine which of the test outputs if any represents an optimal solution.
  • Selection of one or more kernels may be adjusted and one or more support vector machines may be retrained and retested.
  • live data is pre-processed and input into the support vector machine comprising the kernel that produced the optimal solution.
  • the live output from the learning machine may then be post-processed as needed to place the output in a format appropriate for interpretation by a human or another computer.
  • FIG. 1 is a flowchart illustrating an exemplary general method for increasing knowledge that may be discovered from data using a learning machine.
  • FIG. 2 is a flowchart illustrating an exemplary method for increasing knowledge that may be discovered from data using a support vector machine.
  • FIG. 3 is a flowchart illustrating an exemplary optimal categorization method that may be used in a stand-alone configuration or in conjunction with a learning machine for pre-processing or post-processing techniques in accordance with an exemplary embodiment of the present invention.
  • FIG. 4 is a functional block diagram illustrating an exemplary operating environment for an embodiment of the present invention.
  • FIG. 5 is a functional block diagram illustrating a hierarchical system of multiple support vector machines.
  • FIG. 6 shows graphs of the results of using RFE.
  • FIG. 7 shows the results of RFE after preprocessing.
  • FIG. 8 shows the results of RFE when training on 100 dense QT_clust clusters.
  • FIG. 9 shows a comparison of feature (gene) selection methods for colon cancer data.
  • FIG. 10 shows the selection of an optimum number of genes for colon cancer data.
  • FIGS. 14a-d are plots of the results of a sparse SVM (a,c) and a classical SVM (b,d).
  • FIG. 16 is a plot of showing the Hamming distance for a dataset having five labels and two possible label sets.
  • FIG. 17 illustrates a simple toy problem with three labels, one of which is associated with all inputs.
  • FIG. 18. is a histogram of error in the binary approach in the toy problem of FIG. 17.
  • FIGS. 19a-c are histograms of the leave-one-out estimate of the Hamming Loss for the Prostate Cancer Database relating to the embodiment of feature section in multi-label cases
  • FIG. 19a is a histogram of the errors for a direct approach
  • FIG. 19b is a histogram of the errors for the binary approach
  • FIG. 19c is a histogram of the errors for the binary approach when the system is forced to output at least one label.
  • FIG. 20a-c are histograms of the leave-one-out estimate of the Hamming Loss for the Prostate Cancer Database relating to the embodiment of feature section in multi-label cases
  • FIG. 20a is a histogram of the errors for the direct approach
  • FIG. 20b is a histogram of the errors for the binary approach
  • FIG. 20c is a histogram of the errors for the binary approach when the system is forced to output at least one label.
  • FIG. 21 shows the distribution of the mistakes using the leave-one-out estimate of the Hamming Loss for the Prostate Cancer Database using 4 labels with
  • FIG. 21a is a histogram of the errors for the direct approach where the value of one bars is 11.
  • FIG. 21b is a histogram of the errors for the binary approach, where the values of the bars are from left to right: 12 and 7.
  • FIGS. 22a and b are plots showing the results of the transductive SVM using the CORR ub feature selection method: FIG. 22a provides the results for 4 to 20 features compared with the inductive SVM method; FIG. 22b. provides the results for the transductive method using 4 to 100 features.
  • FIGS. 23 a and b are plots showing the results of the transductive CORR ⁇ 2 method compared to the inductive version of the CORR ub 2 method, with FIG. 23 a showing the results or 4 to 25 features and FIG. 23b the results for 4 to 100 features.
  • FIG. 24 is a scatter plot of the seven tumors in the two first gene principal components in an analysis of renal cancer.
  • FIG. 25 is a graph of the distribution of margin values for 4 samples of one class drawn at random according to N(0,1) and 3 samples of another class drawn at random according to N(0,1).
  • FIG. 26 is a plot of the criteria for gene ranking, where FIG. 26a illustrates results for a typical sample drawn from well-separated classes and FIG. 26b illustrates results for a model of an insignificant gene by randomly drawing examples of both classes from the same distribution N(0,1).
  • FIG. 27 is a plot of genes called "significant" versus estimated falsely significant genes for comparing criteria of gene ranking.
  • FIG. 28 is a pair of graphs showing results of SF-SVM analysis of expression data for two genes potentially related to the diseases, selected using the multiclass method, where FIG. 28a is the plot for small inducible cytokine A2 (monocyte chemotactic protein 1, homologous to mounse Sig-je) and FIG. 28b is the plot for ATP synthase, H+ transporting, mitochondrial FI complex, alpha subunit, isoform 1, cardiac muscle.
  • cytokine A2 monoocyte chemotactic protein 1, homologous to mounse Sig-je
  • FIG. 28b is the plot for ATP synthase, H+ transporting, mitochondrial FI complex, alpha subunit, isoform 1, cardiac muscle.
  • the present invention provides methods, systems and devices for discovering knowledge from data using learning machines.
  • the present invention is directed to methods, systems and devices for knowledge discovery from data using learning machines that are provided information regarding changes in biological systems. More particularly, the present invention comprises methods of use of such knowledge for diagnosing and prognosing changes in biological systems such as diseases. Additionally, the present invention comprises methods, compositions and devices for applying such knowledge to the testing and treating of individuals with changes in their individual biological systems. Preferred embodiments comprise detection of genes involved with prostate cancer and use of such information for treatment of patients with prostate cancer.
  • biological data means any data derived from measuring biological conditions of human, animals or other biological organisms including microorganisms, viruses, plants and other living organisms. The measurements may be made by any tests, assays or observations that are known to physicians, scientists, diagnosticians, or the like. Biological data may include, but is not limited to, clinical tests and observations, physical and chemical measurements, genomic determinations, proteomic determinations, drug levels, hormonal and immunological tests, neurochemical or neurophysical measurements, mineral and vitamin level determinations, genetic and familial histories, and other determinations that may give insight into the state of the individual or individuals that are undergoing testing.
  • data is used interchangeably with “biological data”. While several examples of learning machines exist and advancements are expected in this field, the exemplary embodiments of the present invention focus on kernel-based learning machines and more particularly on the support vector machine.
  • the present invention can be used to analyze biological data generated at multiple stages of investigation into biological functions, and further, to integrate the different kinds of data for novel diagnostic and prognostic determinations.
  • biological data obtained from clinical case information such as diagnostic test data, family or genetic histories, prior or current medical treatments, and the clinical outcomes of such activities, can be utilized in the methods, systems and devices of the present invention.
  • clinical samples such as diseased tissues or fluids, and normal tissues and fluids, and cell separations can provide biological data that can be utilized by the current invention.
  • Proteomic determinations such as 2-D gel, mass spectrophotometry and antibody screening can be used to establish databases that can be utilized by the present invention.
  • Genomic databases can also be used alone or in combination with the above-described data and databases by the present invention to provide comprehensive diagnosis, prognosis or predictive capabilities to the user of the present invention.
  • a first aspect of the present invention facilitates analysis of data by pre- processing the data prior to using the data to train a learning machine and/or optionally post-processing the output from a learning machine.
  • pre-processing data comprises reformatting or augmenting the data in order to allow the learning machine to be applied most advantageously. More specifically, pre-processing involves selecting a method for reducing the dimensionality of the feature space, i.e., selecting the features which best represent the data. Methods which may be used for this purpose include recursive feature elimination (RFE), minimizing the number of non-zero parameters of the system, evaluation of cost function to identify a subset of features that are compatible with constraints imposed by the learning set, unbalanced correlation score, inductive feature selection and transductive feature selection. The features remaining after feature selection are then used to train a learning machine for purposes of pattern classification, regression, clustering and/or novelty detection.
  • RFE recursive feature elimination
  • FIG. 1 is a flowchart illustrating a general method 100 for analyzing data using learning machines. The method 100 begins at starting block 101 and progresses to step 102 where a specific problem is formalized for application of analysis through machine learning. Particularly important is a proper formulation of the desired output of the learning machine.
  • a learning machine in predicting future performance of an individual equity instrument, or a market index, a learning machine is likely to achieve better performance when predicting the expected future change rather than predicting the future price level.
  • the future price expectation can later be derived in a post-processing step as will be discussed later in this specification.
  • step 103 addresses training data collection.
  • Training data comprises a set of data points having known characteristics. This data may come from customers, research facilities, academic institutions, national laboratories, commercial entities or other public or confidential sources. The source of the data and the types of data provided are not crucial to the methods. Training data may be collected from one or more local and/or remote sources. The data may be provided through any means such as via the internet, server linkages or discs, CD/ROMs, DVDs or other storage means. The collection of training data may be accomplished manually or by way of an automated process, such as known electronic data transfer methods. Accordingly, an exemplary embodiment of the learning machine for use in conjunction with the present invention may be implemented in a networked computer environment.
  • the collected training data is optionally pre-processed in order to allow the learning machine to be applied most advantageously toward extraction of the knowledge inherent to the training data.
  • this preprocessing stage a variety of different transformations can be performed on the data to enhance its usefulness. Such transformations, examples of which include addition of expert information, labeling, binary conversion, Fourier transformations, etc., will be readily apparent to those of skill in the art.
  • the preprocessing of interest in the present invention is the reduction of dimensionality by way of feature selection, different methods of which are described in detail below.
  • an exemplary method 100 continues at step 106, where the learning machine is trained using the pre-processed data.
  • a learning machine is trained by adjusting its operating parameters until a desirable training output is achieved. The determination of whether a training output is desirable may be accomplished either manually or automatically by comparing the training output to the known characteristics of the training data. A learning machine is considered to be trained when its training output is within a predetermined error threshold from the known characteristics of the training data. In certain situations, it may be desirable, if not necessary, to post-process the training output of the learning machine at step 107. As mentioned, post- processing the output of a learning machine involves interpreting the output into a meaningful form.
  • test data is optionally collected in preparation for testing the trained learning machine.
  • Test data may be collected from one or more local and/or remote sources.
  • test data and training data may be collected from the same source(s) at the same time.
  • test data and training data sets can be divided out of a common data set and stored in a local storage medium for use as different input data sets for a learning machine.
  • any test data used must be pre-processed at step 110 in the same manner as was the training data.
  • a proper test of the learning may only be accomplished by using testing data of the same format as the training data.
  • the learning machine is tested using the pre-processed test data, if any.
  • the test output of the learning machine is optionally post-processed at step 114 in order to determine if the results are desirable.
  • the post processing step involves interpreting the test output into a meaningful form.
  • the meaningful form may be one that is readily understood by a human or one that is compatible with another processor. Regardless, the test output must be post-processed into a form which may be compared to the test data to determine whether the results were desirable. Examples of post-processing steps include but are not limited of the following: optimal categorization determinations, scaling techniques (linear and non-linear), transformations (linear and non-linear), and probability estimations.
  • the method 100 ends at step 116.
  • FIG. 2 is a flow chart illustrating an exemplary method 200 for enhancing knowledge that may be discovered from data using a specific type of learning machine known as a support vector machine (SVM).
  • SVM implements a specialized algorithm for providing generalization when estimating a multidimensional function from a limited collection of data.
  • a SVM may be particularly useful in solving dependency estimation problems. More specifically, a SVM may be used accurately in estimating indicator functions (e.g. pattern recognition problems) and real-valued functions (e.g. function approximation problems, regression estimation problems, density estimation problems, and solving inverse problems).
  • the SVM was originally developed by Vladimir N. Vapnik. The concepts underlying the SVM are explained in detail in his book, entitled Statistical Leaning Theory (John Wiley & Sons, Inc. 1998), which is herein incorporated by reference in its entirety. Accordingly, a familiarity with SVMs and the terminology used therewith are presumed throughout this specification.
  • the exemplary method 200 begins at starting block 201 and advances to step 202, where a problem is formulated and then to step 203, where a training data set is collected.
  • training data may be collected from one or more local and/or remote sources, through a manual or automated process.
  • the training data is optionally pre-processed.
  • pre-processing includes the use of feature selection methods to reduce the dimensionality of feature space.
  • a kernel is selected for the SVM.
  • different kernels will cause a SVM to produce varying degrees of quality in the output for a given set of input data. Therefore, the selection of an appropriate kernel may be essential to the desired quality of the output of the SVM.
  • a kernel may be chosen based on prior performance knowledge.
  • exemplary kernels include polynomial kernels, radial basis classifier kernels, linear kernels, etc.
  • a customized kernel may be created that is specific to a particular problem or type of data set.
  • the multiple SVMs may be trained and tested simultaneously, each using a different kernel.
  • the quality of the outputs for each simultaneously trained and tested SVM may be compared using a variety of selectable or weighted metrics (see step 222) to determine the most desirable kernel.
  • selectable or weighted metrics see step 222
  • locational kernels are defined to exploit patterns within the structure. The locational kernels are then used to construct kernels on the structured object.
  • the pre-processed training data is input into the SVM.
  • the SVM is trained using the pre-processed training data to generate an optimal hyperplane.
  • the training output of the SVM may then be post-processed at step 211.
  • post-processing of training output may be desirable, or even necessary, at this point in order to properly calculate ranges or categories for the output.
  • test data is collected similarly to previous descriptions of data collection.
  • the test data is pre-processed at step 214 in the same manner as was the training data above.
  • the pre-processed test data is input into the SVM for processing in order to determine whether the SVM was trained in a desirable manner.
  • the test output is received from the SVM at step 218 and is optionally post-processed at step 220.
  • an optimal minimum was achieved by the SVM.
  • a SVM is operable to ascertain an output having a global minimum error.
  • output results of a SVM for a given data set will typically vary with kernel selection. Therefore, there are in fact multiple global minimums that may be ascertained by a SVM for a given set of data.
  • the term "optimal minimum” or "optimal solution” refers to a selected global minimum that is considered to be optimal (e.g. the optimal solution for a given set of problem specific, pre-established criteria) when compared to other global minimums ascertained by a SVM.
  • determining whether the optimal minimum has been ascertained may involve comparing the output of a SVM with a historical or predetermined value.
  • a predetermined value may be dependant on the test data set. For example, in the context of a pattern recognition problem where data points are classified by a SVM as either having a certain characteristic or not having the characteristic, a global minimum error of 50% would not be optimal. In this example, a global minimum of 50% is no better than the result that would be achieved by flipping a coin to determine whether the data point had that characteristic.
  • the outputs for each SVM may be compared with output of other SVM to determine the practical optimal solution for that particular set of kernels. The determination of whether an optimal solution has been ascertained may be performed manually or through an automated comparison process.
  • step 224 the kernel selection is adjusted. Adjustment of the kernel selection may comprise selecting one or more new kernels or adjusting kernel parameters. Furthermore, in the case where multiple SVMs were trained and tested simultaneously, selected kernels may be replaced or modified while other kernels may be re-used for control purposes.
  • the method 200 is repeated from step 208, where the pre-processed training data is input into the SVM for training purposes.
  • step 226 live data is collected similarly as described above. By definition, live data has not been previously evaluated, so that the desired output characteristics that were known with respect to the training data and the test data are not known.
  • the live data is pre-processed in the same manner as was the training data and the test data.
  • the live pre-processed data is input into the SVM for processing.
  • the live output of the SVM is received at step 232 and is post-processed at step 234.
  • the method 200 ends at step 236.
  • FIG. 3 is a flow chart illustrating an exemplary optimal categorization method 300 that may be used for pre-processing data or post-processing output from a learning machine. Additionally, as will be described below, the exemplary optimal categorization method may be used as a stand-alone categorization technique, independent from learning machines.
  • the exemplary optimal categorization method 300 begins at starting block 301 and progresses to step 302, where an input data set is received.
  • the input data set comprises a sequence of data samples from a continuous variable.
  • the data samples fall within two or more classification categories.
  • step 304 the bin and class- tracking variables are initialized. As is known in the art, bin variables relate to resolution, while class-tracking variables relate to the number of classifications within the data set.
  • Determining the values for initialization of the bin and class- tracking variables may be perfomied manually or through an automated process, such as a computer program for analyzing the input data set.
  • the data entropy for each bin is calculated.
  • Entropy is a mathematical quantity that measures the uncertainty of a random distribution.
  • entropy is used to gauge the gradations of the input variable so that maximum classification capability is achieved.
  • the method 300 produces a series of "cuts" on the continuous variable, such that the continuous variable may be divided into discrete categories. The cuts selected by the exemplary method 300 are optimal in the sense that the average entropy of each resulting discrete category is minimized.
  • step 308 a determination is made as to whether all cuts have been placed within input data set comprising the continuous variable. If all cuts have not been placed, sequential bin combinations are tested for cutoff determination at step 310. From step 310, the exemplary method 300 loops back through step 306 and returns to step 308 where it is again determined whether all cuts have been placed within input data set comprising the continuous variable. When all cuts have been placed, the entropy for the entire system is evaluated at step 309 and compared to previous results from testing more or fewer cuts. If it cannot be concluded that a minimum entropy state has been determined, then other possible cut selections must be evaluated and the method proceeds to step 311.
  • a heretofore untested selection for number of cuts is chosen and the above process is repeated from step 304.
  • the optimal classification criteria is output at step 312 and the exemplary optimal categorization method 300 ends at step 314.
  • the optimal categorization method 300 takes advantage of dynamic programming techniques. As is known in the art, dynamic programming techniques may be used to significantly improve the efficiency of solving certain complex problems through carefully structuring an algorithm to reduce redundant calculations. In the optimal categorization problem, the straightforward approach of exhaustively searching through all possible cuts in the continuous variable data would result in an algorithm of exponential complexity and would render the problem intractable for even moderate sized inputs.
  • the problem may be divide into a series of sub-problems.
  • algorithmic sub-structures for solving each sub-problem and storing the solutions of the sub-problems, a significant amount of redundant computation may be identified and avoided.
  • the exemplary optimal categorization method 300 may be implemented as an algorithm having a polynomial complexity, which may be used to solve large sized problems.
  • the exemplary optimal categorization method 300 may be used in pre-processing data and/or post-processing the output of a learning machine. For example, as a pre-processing transformation step, the exemplary optimal categorization method 300 may be used to extract classification information from raw data. As a post-processing technique, the exemplary optimal range categorization method may be used to determine the optimal cut-off values for markers objectively based on data, rather than relying on ad hoc approaches. As should be apparent, the exemplary optimal categorization method 300 has applications in pattern recognition, classification, regression problems, etc. The exemplary optimal categorization method 300 may also be used as a stand-alone categorization technique, independent from SVMs and other learning machines.
  • FIG. 4 and the following discussion are intended to provide a brief and general description of a suitable computing environment for implementing biological data analysis according to the present invention.
  • the computer 1000 includes a central processing unit 1022, a system memory 1020, and an Input/Output ("I/O") bus 1026.
  • a system bus 1021 couples the central processing unit 1022 to the system memory 1020.
  • a bus controller 1023 controls the flow of data on the I/O bus 1026 and between the central processing unit 1022 and a variety of internal and external I/O devices.
  • the I/O devices connected to the I/O bus 1026 may have direct access to the system memory 1020 using a Direct Memory Access ("DMA") controller 1024.
  • the I/O devices are connected to the I/O bus 1026 via a set of device interfaces.
  • the device interfaces may include both hardware components and software components. For instance, a hard disk drive 1030 and a floppy disk drive 1032 for reading or writing removable media 1050 may be connected to the I/O bus 1026 through disk drive controllers 1040.
  • An optical disk drive 1034 for reading or writing optical media 1052 may be connected to the I/O bus 1026 using a Small Computer System Interface ("SCSI”) 1041.
  • SCSI Small Computer System Interface
  • an IDE Integrated Drive Electronics, i.e., a hard disk drive interface for PCs
  • ATAPI ATtAchment Packet Interface, i.e., CD-ROM and tape drive interface
  • EIDE Enhanced IDE
  • the drives and their associated computer- readable media provide nonvolatile storage for the computer 1000.
  • other types of computer-readable media may also be used, such as ZIP drives, or the like.
  • a display device 1053 such as a monitor, is connected to the I O bus
  • a parallel interface 1043 connects synchronous peripheral devices, such as a laser printer 1056, to the I/O bus 1026.
  • a serial interface 1044 connects communication devices to the I/O bus 1026.
  • a user may enter commands and information into the computer 1000 via the serial interface 1044 or by using an input device, such as a keyboard 1038, a mouse 1036 or a modem 1057.
  • Other peripheral devices may also be connected to the computer 1000, such as audio input/output devices or image capture devices.
  • a number of program modules may be stored on the drives and in the system memory 1020.
  • the system memory 1020 can include both Random Access Memory (“RAM”) and Read Only Memory (“ROM”).
  • the program modules control how the computer 1000 functions and interacts with the user, with I O devices or with other computers.
  • Program modules include routines, operating systems 1065, application programs, data structures, and other software or firmware components, hi an illustrative embodiment, the learning machine may comprise one or more pre-processing program modules 1075 A, one or more post-processing program modules 1075B, and/or one or more optimal categorization program modules 1077 and one or more SVM program modules 1070 stored on the drives or in the system memory 1020 of the computer 1000.
  • pre-processing program modules 1075 A, post-processing program modules 1075B, together with the SVM program modules 1070 may comprise computer-executable instructions for pre-processing data and post-processing output from a learning machine and implementing the learning algorithm according to the exemplary methods described with reference to FIGS. 1 and 2.
  • optimal categorization program modules 1077 may comprise computer-executable instructions for optimally categorizing a data set according to the exemplary methods described with reference to FIG. 3.
  • the computer 1000 may operate in a networked environment using logical connections to one or more remote computers, such as remote computer 1060.
  • the remote computer 1060 may be a server, a router, a peer device or other common network node, and typically includes many or all of the elements described in connection with the computer 1000.
  • h a networked environment, program modules and data may be stored on the remote computer 1060.
  • the logical connections depicted in FIG. 4 include a local area network (“LAN”) 1054 and a wide area network (“WAN”) 1055.
  • LAN local area network
  • WAN wide area network
  • a network interface 1045 such as an Ethernet adapter card, can be used to connect the computer 1000 to the remote computer 1060.
  • the computer 1000 may use a telecommunications device, such as a modem 1057, to establish a connection.
  • a telecommunications device such as a modem 1057
  • the network connections shown are illustrative and other devices of establishing a communications link between the computers may be used.
  • a plurality of SVMs can be configured to hierarchically process multiple data sets in parallel or sequentially.
  • one or more first-level SVMs may be trained and tested to process a first type of data and one or more first-level SVMs can be trained and tested to process a second type of data. Additional types of data may be processed by other first- level SVMs.
  • the output from some or all of the first-level SVMs may be combined in a logical manner to produce an input data set for one or more second-level SVMs.
  • output from a plurality of second-level SVMs may be combined in a logical manner to produce input data for one or more third-level SVM.
  • the hierarchy of SVMs may be expanded to any number of levels as may be appropriate. In this manner, lower hierarchical level SVMs may be used to pre-process data that is to be input into higher level SVMs. Also, higher hierarchical level SVMs may be used to post-process data that is output from lower hierarchical level SVMs.
  • Each SVM in the hierarchy or each hierarchical level of SVMs may be configured with a distinct kernel.
  • SVMs used to process a first type of data may be configured with a first type of kernel while SVMs used to process a second type of data may utilize a second, different type of kernel.
  • multiple SVMs in the same or different hierarchical level may be configured to process the same type of data using distinct kernels.
  • FIG. 5 illustrates an exemplary hierarchical system of SVMs.
  • one or more first-level SVMs 1302a and 1302b may be trained and tested to process a first type of input data 1304a, such as mammography data, pertaining to a sample of medical patients.
  • a first type of input data 1304a such as mammography data
  • One or more of these SVMs may comprise a distinct kernel, indicated as "KERNEL 1" and "KERNEL 2”.
  • one or more additional first-level SVMs 1302c and 1302d maybe trained and tested to process a second type of data 1304b, which may be, for example, genomic data for the same or a different sample of medical patients.
  • one or more of the additional SVMs may comprise a distinct kernel, indicated as "KERNEL 1" and "KERNEL 3".
  • the output from each of the like first-level SVMs may be compared with each other, e.g., 1306a compared with 1306b; 1306c compared with 1306d, in order to determine optimal outputs 1308a and 1308b.
  • the optimal outputs from the two groups or first-level SVMs i.e., outputs 1308a and 1308b, may be combined to form a new multi-dimensional input data set 1310, for example, relating to mammography and genomic data.
  • the new data set may then be processed by one or more appropriately trained and tested second-level SVMs 1312a and 1312b.
  • resulting outputs 1314a and 1314b from second- level SVMs 1312a and 1312b may be compared to determine an optimal output 1316.
  • Optimal output 1316 may identify causal relationships between the mammography and genomic data points.
  • other combinations of hierarchical SVMs may be used to process either in parallel or serially, data of different types in any field or industry in which analysis of data is desired.
  • Feature Selection by Recursive Feature Elimination The problem of selection of a small amount of data from a large data source, such as a gene subset from a microarray, is particularly solved using the methods, devices and systems described herein. Previous attempts to address this problem used correlation techniques, i.e., assigning a coefficient to the strength of association between variables.
  • correlation techniques i.e., assigning a coefficient to the strength of association between variables.
  • support vector machines methods based on recursive feature elimination (RFE) are used, h examining genetic data to find determinative genes, these methods eliminate gene redundancy automatically and yield better and more compact gene subsets.
  • RFE recursive feature elimination
  • the methods, devices and systems described herein can be used with publicly- available data to find relevant answers, such as genes detenninative of a cancer diagnosis, or with specifically generated data.
  • the illustrative examples are directed at gene expression data manipulations, however, any data can be used in the methods, systems and devices described herein.
  • gene clusters discovered by unsupervised or supervised learning techniques.
  • Preferred methods comprise application of SVMs in determining a small subset of highly discriminant genes that can be used to build very reliable cancer classifiers. Identification of discriminant genes is beneficial in confirming recent discoveries in research or in suggesting avenues for research or treatment. Diagnostic tests that measure the abundance of a given protein in bodily fluids may be derived from the discovery of a small subset of discriminant genes.
  • the input is a vector referred to as a "pattern" of n components referred to as "features”.
  • F is defined as the n- dimensional feature space.
  • the features are gene expression coefficients and the patterns correspond to patients. While the present discussion is directed to two-class classification problems, this is not to limit the scope of the invention.
  • the two classes are identified with the symbols (+) and (-).
  • a training set of a number of patterns ⁇ i, x ,.... ; c , ....xe ⁇ with known class labels y> ⁇ , y ⁇ , ...yu, ....ye ⁇ , yu ⁇ ⁇ -1 5 +1 ⁇ 3 is given.
  • D(x) w x + b, (1) where w is the weight vector and b is a bias value.
  • a data set is said to be linearly separable if a linear discriminant function can separate it without error.
  • Feature selection in large dimensional input spaces is performed using greedy algorithms and feature ranking.
  • a fixed number of top ranked features may be selected for further analysis or to design a classifier.
  • a threshold can be set on the ranking criterion. Only the features whose criterion exceed the threshold are retained.
  • a preferred method uses the ranking to define nested subsets of features, F ⁇ c E 2 c ⁇ • • c F, and select an optimum subset of features with a model selection criterion by varying a single parameter: the number of features.
  • Errorless separation can be achieved with any number of genes greater than one.
  • Preferred methods comprise use of a smaller number of genes.
  • Classical gene selection methods select the genes that individually best classify the training data. These methods include correlation methods and expression ratio methods. While the classical methods eliminate genes that are useless for discrimination (noise), they do not yield compact gene sets because genes are redundant. Moreover, complementary genes that individually do not separate well are missed.
  • a simple feature ranking can be produced by evaluating how well an individual feature contributes to the separation (e.g. cancer vs. nonnal).
  • Various correlation coefficients have been proposed as ranking criteria. For example, see, T.K. Golub, et al, "Molecular classification of cancer: Class discovery and class prediction by gene expression monitoring", Science 286, 531-37 (1999).
  • Each coefficient w t is computed with information about a single feature (gene) and does not take into account mutual information b etween features .
  • One use of feature ranking is in the design of a class predictor (or classifier) based on a pre-selected subset of genes.
  • Each feature which is coreelated (or anti-correlated) with the separation of interest is by itself such a class predictor, albeit an imperfect one.
  • a simple method of classification comprises a method based on weighted voting: the features vote in proportion to their correlation coefficient. Such is the method used by Golub, et al. The weighted voting scheme yields a particular linear discriminant classifier:
  • One aspect of the present invention comprises using the feature ranking coefficients as classifier weights.
  • the weights multiplying the inputs of a given classifier can be used as feature ranking coefficients.
  • the inputs that are weighted by the largest values have the most influence in the classification decision. Therefore, if the classifier performs well, those inputs with largest weights conespond to the most informative features, or in this instance, genes.
  • Other methods known as multivariate classifiers, comprise algorithms to train linear discriminant functions that provide superior feature ranking compared to correlation coefficients. Multivariate classifiers, such as the Fisher's linear discriminant (a combination of multiple univariate classifiers) and methods disclosed herein, are optimized during training to handle multiple variables or features simultaneously.
  • the ideal objective function is the expected value of the enor, i.e., the enor rate computed on an infinite number of examples.
  • this ideal objective is replaced by a cost function J computed on training examples only.
  • Such a cost function is usually a bound or an approximation of the ideal objective, selected for convenience and efficiency.
  • the cost function is:
  • the criteria (w,) 2 estimates the effect on the objective (cost) function of removing feature i.
  • a good feature ranking criterion is not necessarily a good criterion for ranking feature subsets.
  • Some criteria estimate the effect on the objective function of removing one feature at a time. These criteria become suboptimal when several features are removed at one time, which is necessary to obtain a small feature subset.
  • Recursive Feature Elimination (RFE) methods can be used to overcome this problem. RFE methods comprise iteratively 1) training the classifier , 2) computing the ranking criterion for all features, and 3) removing the feature having the smallest ranking criterion. This iterative procedure is an example of backward feature elimination.
  • the weights of a classifier are used to produce a feature ranking with a SVM (Support Vector Machine).
  • SVM Small Vector Machine
  • the present invention contemplates methods of SVMs used for both linear and non-linear decision boundaries of arbitrary complexity, however, the example provided herein is directed to linear SVMs because of the nature of the data set under investigation.
  • Linear SVMs are particular linear discriminant classifiers. (See Equation 1). If the training set is linearly separable, a linear SVM is a maximum margin classifier. The decision boundary (a straight line in the case of a two-dimension separation) is positioned to leave the largest possible margin on either side.
  • weights w,- of the decision function D(x) are a function only of a small subset of the training examples, i.e., "support vectors".
  • Support vectors are the examples that are closest to the decision boundary and lie on the margin. The existence of such support vectors is at the origin of the computational properties of SVM and its competitive classification performance. While SVMs base their decision function on the support vectors that are the borderline cases, other methods such as the previously-described method of Golub, et al., base the decision function on the average case.
  • a preferred method of the present invention comprises using a variant of the soft-margin algorithm where training comprises executing a quadratic program as described by Cortes and Vapnik in "Support vector networks", 1995, Machine Learning, 20:3, 273-297, which is incorporated herein by reference in its entirety.
  • training comprises executing a quadratic program as described by Cortes and Vapnik in "Support vector networks", 1995, Machine Learning, 20:3, 273-297, which is incorporated herein by reference in its entirety.
  • the following is provided as an example, however, different programs are contemplated by the present invention and can be determined by those skilled in the art for the particular data sets involved.
  • Inputs comprise training examples (vectors) ⁇ x 1 ⁇ X ⁇ ,....x k ...x e ⁇ and class labels ⁇ y ls y 2 ....y/ c ...y* ⁇ .
  • the soft margin parameters ensure convergence even when the problem is non-linearly separable or poorly conditioned. In such cases, some support vectors may not lie on the margin.
  • the weight vector w is a linear combination of training patterns. Most weights k are zero.
  • the training patterns with non-zero weights are support vectors. Those having a weight that satisfies the strict inequality 0 ⁇ a ⁇ C are marginal support vectors.
  • the bias value b is an average over marginal support vectors.
  • the following sequence illustrates application of recursive feature elimination (RFE) to a SVM using the weight magnitude as the ranking criterion.
  • the output comprises feature ranked list r.
  • the above steps can be modified to increase computing speed by generalizing the algorithm to remove more than one feature per step.
  • RFE is computationally expensive when compared against correlation methods, where several thousands of input data points can be ranked in about one second using a Pentium® processor, and weights of the classifier trained only once with all features, such as SVMs or pseudo-inverse/mean squared error (MSE).
  • SVMs implemented using non-optimized MatLab® code on a Pentium® processor can provide a solution in a few seconds.
  • MSE pseudo-inverse/mean squared error
  • a SVM implemented using non-optimized MatLab® code on a Pentium® processor can provide a solution in a few seconds.
  • RFE is preferrably implemented by training multiple classifiers on subsets of features of decreasing size. Training time scales linearly with the number of classifiers to be trained. The trade-off is computational time versus accuracy. Use of RFE provides better feature selection than can be obtained by using the weights of a single classifier.
  • RFE can be used by removing chunks of features in the first few iterations and then, in later iterations, removing one feature at a time once the feature set reaches a few hundreds.
  • RFE can be used when the number of features, e.g., genes, is increased to millions, hi one example, at the first iteration, the number of genes were reached that was the closest power of two. At subsequent iterations, half of the remaining genes were eliminated, such that each iteration was reduced by a power of two. Nested subsets of genes were obtained that had increasing information density.
  • RFE consistently outperforms the na ⁇ ve ranking, particularly for small feature subsets.
  • the na ⁇ ve ranking comprises ranking the features with (w,) 2 , which is computationally equivalent to the first iteration of RFE.
  • the na ⁇ ve ranking organizes features according to their individual relevance, while RFE ranking is a feature subset ranking.
  • the nested feature subsets contain complementary features that individually are not necessarily the most relevant.
  • An important aspect of SVM feature selection is that clean data is most preferred because outliers play an essential role. The selection of useful patterns, support vectors, and selection of useful features are connected.
  • Pre-processing can have a strong impact on SVM-RFE.
  • feature scales must be comparable.
  • One pre-processing method is to subtract the mean of a feature from each feature, then divide the result by its standard deviation. Such pre-processing is not necessary if scaling is talcen into account in the computational cost function.
  • Another pre-processing operation can be performed to reduce skew in the data distribution and provide more uniform distribution. This pre-processing step involves taking the log of the value, which is particularly advantageous when the data consists of gene expression coefficients, which are often obtained by computing the ratio of two values. For example, in a competitive hybridization scheme, DNA from two samples that are labeled differently are hybridized onto the anay.
  • the first initial preprocessing step that is taken is to take the ratio ab of these two values. Though this initial preprocessing step is adequate, it may not be optimal when the two values are small.
  • Other initial preprocessing steps include (a-b)/(a+b) and (log a - log b)/(log a + log b).
  • Another pre-processing step involved normalizing the data across all samples by subtracting the mean.
  • This preprocessing step is supported by the fact that, using tissue samples, there are variations in experimental conditions from microarray to microarray. Although standard deviation seems to remain fairly constant, the other preprocessing step selected was to divide the gene expression values by the standard deviation to obtain centered data of standardized variance. To normalize each gene expression across multiple tissue samples, the mean expression value and standard deviation for each gene was computed. For all the tissue sample values of that gene (training and test), that mean was then subtracted and the resultant value was divided by the standard deviation.
  • the data can be pre-processed by a simple "whitening" to make data matrix resemble "white noise.”
  • the samples can be pre-processed to: nonnalize matrix columns; normalize matrix lines; and normalize columns again. Normalization consists of subtracting the mean and dividing by the standard deviation.
  • a further normalization step can be taken when the samples are split into a training set and a test set. The mean and variance column-wise was computed for the training samples only.
  • SVM-RFE can be used in nonlinear cases and other kernel methods.
  • the method of eliminating features on the basis of the smallest change in cost function can be extended to nonlinear uses and to all kernel methods in general. Computations can be made tractable by assuming no change in the value of the ⁇ 's. Thus, the classifer need not be retrained for every candidate feature to be eliminated. Specifically, in the case of SVMs, the cost function to be minimized
  • H is the matrix with elements ⁇ l ⁇ k K x. ⁇ l x k ), K is a kernel function that measures the similarity between x /; and x admir and 1 is an 2 dimensional vector of ones.
  • EXAMPLE 1 Analysis of gene patterns related to colon cancer Analysis of data from diagnostic genetic testing, microarray data of gene expression vectors, was performed with a SVM-RFE. The original data for this example was derived from the data presented in Alon et al., 1999. Gene expression information was extracted from microanay data resulting, after preprocessing, in a table of 62 tissues x 2000 genes. The 62 tissues include 22 normal tissues and 40 colon cancer tissues. The matrix contains the expression of the 2000 genes with highest minimal intensity across the 62 tissues. Some of the genes are non-human genes.
  • the data proved to be relatively easy to separate. After preprocessing, it was possible to a find a weighted sum of a set of only a few genes that separated without error the entire data set, thus the data set was linearly separable.
  • One problem in the colon cancer data set was that tumor samples and normal samples differed in cell composition. Tumor samples were normally rich in epithelial cells wherein normal samples were a mixture of cell types, including a large fraction of smooth muscle cells. While the samples could be easily separated on the basis of cell composition, this separation was not very infonnative for tracking cancer-related genes.
  • the gene selection method using RFE-SVM is compared against a reference gene selection method described in Golub et al, Science,1999, which is referred to as the "baseline method" Since there was no defined training and test set, the data was randomly split into 31 samples for training and 31 samples for testing.
  • the authors use several metrics of classifier quality, including enor rate, rejection rate at fixed threshold, and classification confidence. Each value is computed both on the independent test set and using the leave-one-out method on the training set.
  • the leave-one-out method consists of removing one example from the training set, constructing the decision function on the basis only of the remaining training data and then testing on the removed example. In this method, one tests all examples of the training data and measures the fraction of enors over the total number of training examples.
  • the methods of this Example uses a modification of the above metrics.
  • the present classification methods use various decision functions (D(x) whose inputs are gene expression coefficients and whose outputs are a signed number indicative of whether or not cancer was present.
  • the classification decision is canied out according to the sign of D(x).
  • the magnitude of D(x) is indicative of classification confidence.
  • E/D difference between the smallest output of the positive class samples and the largest output of the negative class samples (rescaled by the largest difference between outputs)
  • Median margin (M/D) difference between the median output of the positive class samples and the median output of the negative class samples (rescaled by the largest difference between outputs).
  • the enor rate is complemented by the success rate.
  • the rejection rate is the fraction of examples that are rejected (on which no decision is made because of low confidence).
  • the rejection rate is complemented by the acceptance rate.
  • Extremal and median margins are measurements of classification confidence. Note that the margin computed with the leave-one-out method or on the test set differs from the margin computed on training examples sometimes used in model selection criteria.
  • a method for predicting the optimum subset of genes comprised defining a criterion of optimality that uses information derived from training examples only. This criterion was checked by determining whether the predicted gene subset performed best on the test set.
  • a criterion that is often used in similar "model selection" problems is the leave-one-out success rate V SU c- h the present example, it was of little use since differentiation between many classifiers that have zero leave-one-out enor is not allowed. Such differentiation is obtained by using a criterion that combines all of the quality metrics computed by cross-validation with the leave-one-out method:
  • Theoretical considerations suggested modification of this criterion to penalize large gene sets.
  • a SVM-RFE was run on the raw data to assess the validity of the method.
  • the colon cancer data samples were split randomly into 31 examples for training and 31 examples for testing.
  • the RFE method was run to progressively downsize the number of genes, each time dividing the number by 2.
  • the pre-processing of the data for each gene expression value consisted of subtracting the mean from the value, then dividing the resultby the standard deviation.
  • the leave-one-out method with the classifier quality criterion was used to estimate the optimum number of genes.
  • the leave-one-out method comprises taking out one example of the training set. Training is then perfonned on the remaining examples, with the left out example being used to test the trained classifier. This procedure is iterated over all the examples. Each criteria is computed as an average over all examples.
  • the overall classifier quality criterion is calculated according to Equation 13.
  • the classifier is a linear classifier with hard margin.
  • the optimum test performance had an 81% success rate without pre- processing to remove skew and to normalize the data. This result was consistent with the results reported in the original paper by Alon et al. Moreover, the enors, except for one, were identified by Alon et al. as outliers.
  • the plot of the performance curves as a function of gene number is shown in FIG. 6.
  • the predictor of optimum test success rate (diamond curve), which is obtained by smoothing after substracting ⁇ from the leave-one-out quality criterion, coincides with the actual test success rate (circle curve) in finding the optimum number of genes to be 4.
  • FIG. 7 shows the results of RFE after preprocessing, where the predicted optimum test success rate is achieved with 16 genes.
  • the reduced capacity SVM used in FIG. 6 is replaced by plain SVM.
  • a log scale is still used for gene number, RFE was run by eliminating one gene at a time.
  • the best test performance is 90% classification accuracy (8 genes).
  • the optimum number of genes predicted from the classifier quality based on training data infonnation only is 16. This corresponds to 87% classification accuracy on the test set.
  • SVM-RFE used a subset of genes that were complementary and thus canied little redundant information. No other information on the structure and nature of the data was provided. Because data were very redundant, a gene that had not been selected may nevertheless be informative for the separation.
  • Conelation methods such as Golub 's method provide a ranked list of genes.
  • the rank order characterizes how conelated the gene is with the separation.
  • a gene highly ranked taken alone provides a better separation than a lower ranked gene. It is therefore possible to set a threshold (e.g. keep only the top ranked genes) that separates "highly informative genes” from “less informative genes”.
  • the methods of the present invention such as SVM-RFE provide subsets of genes that are both smaller and more discriminant.
  • the gene selection method using SVM-RFE also provides a ranked list of genes. With this list, nested subsets of genes of increasing sizes can be defined. However, the fact that one gene has a higher rank than another gene does not mean that this one factor alone characterizes the better separation. In fact, genes that are eliminated in an early iteration could well be very informative but redundant with others that were kept.
  • the 32 best genes as a whole provide a good separation but individually may not be very conelated with the target separation.
  • Gene ranking allows for a building nested subsets of genes that provide good separations, however it provides no information as to how good an individual gene may be. Genes of any rank may be correlated with the 32 best genes.
  • the conelated genes may be ruled out at some point because of their redundancy with some of the remaining genes, not because they did not carry infonnation relative to the target separation.
  • the gene ranking alone is insufficient to characterize which genes are informative and which ones are not, and also to determine which genes are complementary and which ones are redundant. Therefore, additional preprocessing in the form of clustering was performed.
  • FIG. 8 provides the performance curves of the results of RFE when trained on 100 dense QT clust clusters. As indicated, the predicted optimum number of gene cluster centers is 8. The results of this analysis are comparable to those of FIG. 7.
  • the cluster centers can be substituted by any of their members. This factor may be important in the design of some medical diagnosis tests. For example, the administration of some proteins may be easier than that of others. Having a choice of alternative genes introduces flexibility in the treatment and administration choices .
  • Hierarchical clustering instead of QT c ⁇ Ust clustering was used to produce lots of small clusters containing 2 elements on average. Because of the smaller cluster cardinality, there were fewer gene alternatives from which to choose, h this instance, hierarchical clustering did not yield as good a result as using QT c ⁇ ust clustering.
  • the present invention contemplates use of any of the known methods for clustering, including but not limited to hierarchical clustering, QT c ⁇ us t clustering and SVM clustering.
  • clustering method to employ in the invention is affected by the initial data and the outcome desired, and can be determined by those skilled in the art.
  • Another method used with the present invention was to use clustering as a post-processing step of SVM-RFE.
  • Each gene selected by rumiing regular SVM- RFE on the original set of gene expression coefficients was used as a cluster center. For example, the results described with reference to FIG. 7 were used.
  • the conelation coefficient was computed with all remaining genes. The parameters were that the genes clustered to gene i were those that met the following two conditions: higher conelation coefficient with gene i than with other genes in the selected subset of eight genes, and conelation coefficient exceeding a threshold ⁇ .
  • the supervised clustering method does not provide better control over the number of examples per cluster. Therefore, this method is not as good as unsupervised clustering if the goal is the ability to select from a variety of genes in each cluster.
  • supervised clustering may show specific clusters that have relevance for the specific knowledge being determined, hi this particular embodiment, in particular, a very large cluster of genes was found that contained several muscle genes that may be related to tissue composition and may not be relevant to the cancer vs. normal separation. Thus, those genes are good candidates for elimination from consideration as having little bearing on the diagnosis or prognosis for colon cancer.
  • tissue composition-related genes were automatically eliminated in the pre-processing step by searching for the phrase "smooth muscle". Other means for searching the data for indicators of the smooth muscle genes may be used.
  • SVM-RFE Linear Discriminant Analysis
  • MSE Mean Squared Enor
  • SVM-RFE provided better performance than the other methods. All methods predicted an optimum number of genes smaller or equal to 64 using the criterion of the Equation 15. The genes ranking 1 through 64 for all the methods studied were compared. The first gene that was related to tissue composition and mentions "smooth muscle" in its description ranks 5 for Golub's method, 4 for LDA, 1 for MSE and only 41 for SVM. Therefore, this was a strong indication that SVMs make a better use of the data compared with other methods since they are the only methods that effectively factors out tissue composition-related genes while providing highly accurate separations with a small subset of genes.
  • FIG. 10 is a plot of an optimum number of genes for evaluation of colon cancer data using RFE-SVM.
  • the number of genes selected by recursive gene elimination with SVMs was varied and a number of quality metrics were evaluated include error rate on the test set, scaled quality criterion (Q/4), scaled criterion of optimality (C/4), locally smoothed C/4 and scaled theoretical enor bar ( ⁇ /2).
  • the model selection criterion was used in a variety of other experiments using SVMs and other algorithms.
  • the optimum number of genes was always predicted accurately, within a factor of two of the number of genes.
  • a second method of feature selection according to the present invention comprises minimizing the ⁇ o-norm of parameter vectors.
  • Such a procedure is central to many tasks in machine learning, including feature selection, vector quantization and compression methods.
  • This method consfructs a classifier which separates data using the smallest possible number of features.
  • the £ 0 -norm of w is minimized by solving the optimization problem minlHlo subject to : ⁇ .((w,x,.) + b)> l , (16) w where (
  • j 0 card ⁇ w,-
  • the goal is to find the fewest non-zero elements in the vector of coefficients w.
  • this problem is combinatorially hard, the following approximation is used:
  • Equation 17 can be solved by using constrained gradient.
  • Equation 20 The second term of the right hand side of Equation 20 is negligible compared to the ln( ⁇ ) term when ⁇ is very small.
  • ALOM solves a succession of linear optimization problems with non- sparse constraints. Sometimes, it may be more advantageous to have a quadratic programming formulation of these problems since the dual may have simple constraints and may then become easy to solve.
  • the present embodiment uses a procedure refened to as "£ 2 - ALOM" to minimize the o norm as follows:
  • This method is developed for a linearly-separable learning set.
  • the vector w c is learned by discriminating the class c from all other classes. This gives many two-class problems. In this framework, the minimization of the £o-norm is done for each vector w c independently of the others. However, the true o-norm is the following:
  • any kind of function k(. , .) can be used instead of ⁇ . , if k can be understood as a dot-product.
  • the function ⁇ (x) ⁇ (x), ..., ⁇ t (x) e 2 which maps points x ; into a feature space such that k(x, x 2 ) can be interpreted as a dot product in feature space.
  • Multiplications in feature space are of the form (x) * ⁇ (y).
  • ⁇ d (x * y) (x ; ⁇ y,_ ... ⁇ ⁇ : i ⁇ i x , i 2 ,..., i d ⁇ N) .
  • ⁇ ⁇ ) ⁇ P (x) ' - 1 ⁇ P ⁇ d) and
  • test points can be classified using
  • the first six features have redundancy and the rest of the features are inelevant.
  • Linear decision rules were used and feature selection was perfonned selecting the two best features using each of the above-mentioned methods along with ALOM SNMs, using both 2 ⁇ and £ 2 multiplicative updates.
  • Training was performed on 10, 20 and 30 randomly drawn training points, testing on a further 500 points, and averaging test enor over 100 trials. The results are provided in Table 1. For each technique, the test enor and standard deviation are given.
  • the Q approximation compared to RFE also has a lower computational cost, hi the RFE approach, n iterations are performed, removing one feature per iteration, where n is the number of input dimensions.
  • RFE can be sped up by removing more than one feature at a time.
  • Table 2 provides the p-values for the null hypothesis that the algorithm in a given row does not outperform the algorithm in a given column using the Wilcoxon signed rank test.
  • the Wilcoxon test evaluates whether the generalization enor of one algorithm is smaller than another. The results are given for 10, 20 and 30 training points.
  • the 2 i-ALOM SVM method outperforms all other methods.
  • the next best results are obtained with the 2 -AL0M SVM.
  • This ranking is consistent with the theoretical analysis of the algorithm — the 2 2 -norm approximation should not be as good at choosing a small subset of features relative to the 2 ⁇ approximation which is closer to the trae fo-nonn minimization.
  • ALOM SVMs slightly outperform RFE SVMs, whereas conelation coefficients (CORR SVM) are significantly worse.
  • Table 4 provides the p-values using the Wilcoxon sign rank test to demonstrate the significance of the difference between algorithms, again showing that RFE SVM and £ 2 -AL0M outperform conelation coefficients. 2 -AL0M outperforms RFE for small feature set sizes.
  • Table 5 RFE and the approximation to the 2 -AL0M again outperform conelation coefficients. 2 -AL0M and RFE provided comparable results. Table 6 gives the p-values using the Wilcoxon sign ranlc test to show the significance of the difference between algorithms.
  • a micro anay dataset of 208 genes (Brown Yeast dataset) was discriminated into five classes based on 79 gene expressions conesponding to different experimental conditions..
  • Two 8 cross-validation runs were performed. The first run was done with a classical multiclass SVM without any feature selection method. The second ran was done with a SVM and a pre-processing step using the multiclass 2 -AL0M procedure to select features. Table 7 shows the results, i.e., that the 2 -AL0M multiclass SVM outperforms the classical multiclass SVM. As indicated, the number of features following feature selection is greatly reduced relative to the original set of features.
  • kernel space feature selection is usually one of improving generalization performance rather than improving running time or attempting to interpret the decision rule.
  • Problems (e) and (f) show an 80% and 0% sparse, respectively. Note that while the method outperforms SVMs in the case of spare targets, it is not much worse when the target is not sparse, as in problem (f). In problem (e), the ALOM method is slightly better than SVMs when there are more than 50 training points, but worse otherwise. This may be due to error when making sparse rules when the data is too scarce. This suggests that it may be preferable in certain cases to choose a rule that is only partially sparse, i.e., something in between the i? 2 -norm and o-norm. It is possible to obtain such a mixture by considering individual iterations.
  • the ⁇ -no ⁇ m classifier separates data using the least possible number of features. This is a combinatorial optimization problem wliich is solved approximately by finding a local minimum through the above-described techniques. These features can then be used for another classifier such as a SVM. If there are still too many features, they can be ranked according to the absolute value of the weight vector of coefficient assigned to them by the separating hyperplane.
  • Feature selection methods can also be applied to define very sparse SVM. For that purpose, it is useful to review the primal optimization problem from which SVMs are defined:
  • the present method tends to behave like a SVM but with a very sparse expansion (see Equation 38).
  • the learning set consists of 100 points in [0,1] 2 with ⁇ 1 targets as they are drawn in FIG. 13.
  • FIG. 13 a the result of the sparse SVM is shown, while FIG. 13b shows the result of a classic SVM.
  • the parse-SVM obtains a sparser and a better solution in terms of generalization enor.
  • the circled points in the plots indicate learning points with a non-zero oti.
  • VQ kernel-based vector quantization
  • VQ tries to represent a set of 2 data vectors x l t . . . , x t e ⁇ (40) by a reduced number of m codebook vectors y ⁇ , . . . y m ⁇ , (41) such that some measure of distortion is minimized when each x is represented by the nearest (in terms of some metric on )y.
  • codebooks are constructed by greedy minimization of the distortion measure, an example being the overall 2_ enor.
  • each x can then be compressed into log 2 m bits.
  • VQ algorithms using multiplicative updates can be obtained according to the following:
  • ⁇ is some space endowed with a metric d.
  • kernel will be used in a broader sense to mean functions
  • the x j with nonzero coefficients Wj can be considered an approximation of the complete training set. (Formally, they define a R-cover of the training set in the metric d.) h order to keep the number of nonzero coefficients small, the following optimization problem is considered: for some q ⁇ 0, compute: minllwl (48)
  • a pruning step is performed to remove some of the remaining redundant vectors in the codebook, by sequentially removing any codebook vector that does not exclusively explain any data vector. Equivalently, the pruning step can be applied to the training set before training. If there are two points whose R-balls cover the same subsets of the training set, they can be considered equivalent, and one of them can be removed. Typically, the pruning results in the removal of a further 1-5% of the codebook.
  • chunking or decomposition methods can be employed. Such methods include (1) where the inner optimization uses the 2 -norm, it has a structure similar to the standard SVM optimization problem. In this case, SVM chunlcing methods can be used. (2) Forms of chunlcing can be derived directly from the problem stracture. These include greedy chunking, SV-style chunlcing, and parallel chunlcing.
  • Greedy chunlcing involves two steps.
  • Step 1 start with a first chunk of 0 ⁇ 2Q ⁇ 2 datapoints. Solve the problem for these points, obtaining a set of m ⁇ £ 0 codebook vectors.
  • Step 2 go through the remaining 2-2Q points and discard all points that have already been covered.
  • Step 3 provided there are still points left, talce a new chunk of 2 from the remaining set, find the codebook vectors, and removed all points which are already covered.
  • SV style chunking has the following inner loop:
  • Step 1 Start with the first chunk of 0 ⁇ 2Q ⁇ 2 datapoints. Solve the problem for these points, obtaining a set of m ⁇ Eo codebook vectors and discarding the rest from the cunent chunk. Next, start going through the remaining 2 - 2Q points and fill up the cunent chunk until is has size 2Q again. In Step n, provided there are still points left and the cunent chunk has size smaller than 2o, proceed as above and fill it up. If it is already full, remove a fixed number of codebook vectors along with all points falling into their respective r- balls.
  • the dataset is split into p parts of equal size.
  • the standard algorithm is run on each part using a kernel of width R/2. Once this is done, the union of all codebooks is fonned and a cover is computed using width R/2. By the triangular inequality, a R-cover of the whole dataset is obtained.
  • Useful heuristics for finding such candidate points include a scheme where, for each codebook vector, a new point is generated by moving the codebook vector in the direction of a point having the property that, among all points that are coded only by the present codebook vector, has the maximum distance to that codebook vector. If the distance of the movement is such that no other points leave the cover, the original codebook vector can be discarded in favor of the new one.
  • the proposed algorithm finds solution which are either optimal or close to optimal. Optimality is assessed using the following greedy covering algorithm. At each step, find a point with the property that the number of points contained in aR-ball around it is maximal. Add this point to the codebook, remove all points in the ball from the training set, and, if there are still points left, go back to the beginning, hi FIGS.
  • the VQ method described herein allows an upper bound to be specified on the distortion incuned for each coded vector.
  • the algorithm automatically determines small codebooks covering the entire dataset. It works on general domains endowed with a metric, and could this equally well be used to compute coverings of function spaces.
  • the algorithm could be used for data reduction as a pre-processing step, e.g., for classification, by separating the codebook vectors with a margin larger than R.
  • Target values can be incorporated in the constraint such that it is preferable for the algorithm to find "pure clusters".
  • the term vector quantization has been used, the method can be applied to non-vectorial data, as long as it is possible to define a suitable distance measure d.
  • VQ methods may also be used to compress or encode images.
  • n- norm for multi-label problems Use of n- norm for multi-label problems.
  • the preceding method for feature selection can be used in cases of multi-label problems such as frequently arise in bioinformatics.
  • the same multiplicative update rale is used with the additional step of computing the label sets size s(x) using ranking techniques.
  • a cost function and margin for multi-label models are defined. Cost functions for multi-label problems are defined as follows:
  • An output space is considered as the space formed by all the sets of integer between 1 and Q identified as the labels of the learning problem. Such an output space contains 2Q elements and one output conesponds to one set of labels.
  • the function c is a real-valued loss and can talce different forms depending on how it is viewed.
  • two types of loss are considered. The first is the "Hamming Loss", which is defined as
  • ⁇ stands for the symmetric difference of sets. Note that the moref(x) is different from Y, the higher the enor. Missing one label in Y is less important than missing two, which seems quite natural in many real situations. As an example, consider that the labels are possible diseases related to different genes. Many genes can share the same disease and lead to many diseases at the same time. Predicting only one disease although there are three is worse than predicting only two. Having a Hamming Loss of 0.1 means that the expected number of times a pair (x, yi has been misclassified is 0.1. Note that if the Hamming Loss is scaled by Q, it is equal to the average number of wrong labels for each point.
  • f(x) 8ig [(wjx) + b],...,(w Q ,x) + b Q ) (54) where the sign function applies component- wise.
  • the value of f(x) is a binary vector from which the set of labels can be retrieved easily by stating that label k is in the set ⁇ (w k ,x) + b k )> 0. This way of computing the output is discussed below.
  • the natural cost function for such a model is the Hamming Loss.
  • Equation 54 the function/ is computed as in Equation 54.
  • the margin of / on (x, Y) is defined as the signed distance between ( ⁇ w ⁇ , >)-l- b ⁇ , .. . , ⁇ W Q , X> + b ⁇ ) and the decision boundary. It is equal to:
  • the margin can also be defined as:
  • Y,- is identified with its binary representation: (yn, .., y t ⁇ ) € ⁇ -1, +1 ⁇ Q . Maximizing the margin or minimizing its inverse yields to the following problem: min maxllw,. I (59)
  • the previous problem can be generalized by combining the minimization of the Hamming Loss and the maximization of the margin:
  • Equation 62 the Hamming Loss is replaced by its linear approximation (here computed on (x, Y)):
  • Equation 65 is rewritten as Equation 54.
  • J _t ⁇ -1, +l ⁇ ⁇ both computations are different (see FIG. 16) and/jc) should be calculated as above rather than using Equation 54.
  • Such cases arise when, e.g., a Enor Correcting Output Code (ECOC) is used to solve multi-class problems.
  • ECOC Enor Correcting Output Code
  • (66) k Y ' where ⁇ is the linear function thresholded at -1 (resp. +1) with value -1 (resp. +1).
  • the ranking approach can be broken down into two parts, the first of which is to optimize the Ranking Loss.
  • the second part is obtained by minimizing the Size Loss 1/Q 1 1 Y
  • This second part is actually almost a regression problem and can be solved with the approaches known in the prior art.
  • the goal is to define a linear model that minimizes the Ranking Loss while having a low complexity.
  • the notion of complexity is the margin.
  • the margin of (x, Y) can be expressed as:
  • the objective function can be replaced by: max min The w ⁇ k, ⁇
  • MLR-SVM Support Vector Machine
  • Categorical regression can be dealt with using the multi-label approach. This approach provides a natural solution of the categorical regression problem when the right setting is used. Rather than coding the labels as integers, encode them as binary vector of ⁇ +1, -1 ⁇ components:
  • the loss CR defined previously can be expressed in terms of the Hamming Loss:
  • the multi-label model is decomponsed into two parts, both based on dot products and linear models.
  • the first part ranks the labels and is obtained via Equation 74.
  • the result is a set of parameters
  • Equation 62 l, . . , Q.
  • Equation 62 l, . . , Q.
  • the second part of the model predicts the size of each label sets and is obtained through Equation 62, where the maximum has been replaced by a sum.
  • the result is also a set of parameters (w k 2 ,
  • the output is computed by talcing the integer whose representation (Equation 76) is the closest from the vector:
  • Equation 74 the dual variables a i ⁇ > 0 are related to the constraints are introduced: ( W/c _ W/ 5 / ) + ⁇ _ ⁇ / _ ⁇ + ⁇ / ⁇ o (78) and the variables ⁇ i ⁇ > 0 related to the constraints ⁇ ik ⁇ > 0.
  • the lagrangian can then be computed:
  • Equation 80 Equation 80
  • Equation 74 The dual of Equation 74 can then be expressed, h order to have as simple notation as possible, it will be expressed with both variables,? / ⁇ - and m.
  • Equation 81 The first box constraints are derived according to Equation 81, by using the fact that rjiki ⁇ 0.
  • the dual problem has an advantage compared to its primal counterpart.
  • the main problem in this case concerns the objective function: it is a quadratic function that cannot be solved directly except by storing its whole Hessian.
  • the number of variables may be too important to take a direct approach. For this reason, an approximation scheme from Franke and Wolfe is followed.
  • ⁇ ⁇ are computed from Equation 83 in terms of the a p t ⁇ .
  • the latter are the cunent values of the parameter optimized using Franke and Wolfe's method.
  • w k v k .
  • the objective function can thus be expressed as:
  • the Jpart can be differentiated directly.
  • the /part can be differentiated as:
  • the computation of the gradient of the objective function can thus be done directly as soon as the vectors V k P are given.
  • the vectors are expressed in terms of the a p ,ki and only dot products between them and the x,'s are needed, which means that this procedure can also be used for kernels.
  • the x- axis represents the number of erroneous labels (missed or added) for the learning set of 50 elements; the y-axis represents the number of points out of 50.
  • the enor for a multi-label system is represented using this charts and the Hamming Loss.
  • Such a representation enables assessment of the quality of a multi-label classifier with respect to many criterion, namely, the number of misclassified points, the number of classes that are badly predicted, etc.
  • the Hamming Loss is proportional to the sum of the height of the bars scaled by their coordinates.
  • the dataset is formed by 67 examples in a space of 7129 features and of 9 labels.
  • the inputs are results of Micro-array experiments on different tissues coming from patients with different form of Prostate Cancer.
  • the nine labels are described in Table 8 and represent either the position of the tissue in the body (peripheral zone, etc..) or the degree of malignity of the disease (G3-G4 Cancer).
  • label 4 has only one point which implies that a direct approach will have a leave-one-out Hamming Loss of at least 0.02 (conesponding to the enor when the points labelled Stroma is out).
  • the LCM label is associated with examples that have been analyzed with a Laser Confocal Microscopy technique, and the 2-amplifications label refer to examples that have had two amplifications during the PCR. These two labels are discarded so that the remaining problem is one of separating different tissue types, thus reducing the number of labels to 7.
  • the label sets have a maximum size of 2.
  • FIG. 19a-c shows the distribution of the enors for the direct and the binary approach in the leave-one-out estimate of the Hamming Loss.
  • FIG. 19a shows the enors for the direct approach, where the values of the bars are from left to right: 4, 19 and 3.
  • FIG. 19b shows the enors for the binary approach where the values of the bars are from left to right: 20, 10 and 1.
  • FIG. 19c shows the enors for the binary approach when the system is forced to output at least one label, where the values of the bars are from left to right: 9, 16 and 2.
  • the direct approach yields mistakes on 26 points although the binary approach makes mistakes on 31 points, h terms of the Hainming Loss, the binary approach is the best although in terms of the number of conectly classified input points, the direct approach wins.
  • the binary approach is naturally related to the Hamming Loss and it seems quite natural that it minimizes its value.
  • the direct approach is better.
  • the multi-labelled approach was applied for labels 4-7 only, reducing the number of labels to 4 and the number of points in the learning set to 56.
  • the leave-one-out estimate of the Hamming Loss was 0.14 for the direct approach, and 0.14 for the binary approach.
  • the same experiment as above was perfonned by computing the Hamming Loss of the binary system when the latter is forced to give at least one label, producmg a Hamming Loss of 0.16. i this particular application, the direct approach is better, yielding a system with the same or a lower Hamming Loss than the binary approach.
  • FIG. 20a-c Prostate Cancer Database using 4 labels are provided in FIG. 20a-c.
  • FIG. 20a is a histogram of the errors for the direct approach, where the values of the bars are from left to right: 2, 13 and 1).
  • FIG. 20b is a histogram of the enors for the binary approach, where the values of the bars are from left to right: 12, 8 and 1.
  • FIG. 20c is a histogram of the enors for the binary approach when the system is forced to output at least one label, where the values of the bars are from left to right: 7, 13 and 1.
  • This method is an approximation scheme to minimize the number of non-zero components of (w ⁇ , .., W ⁇ ) while keeping the constraints of
  • Equation 92 Equation 92 satisfied. It is shown to converge in the prior discussion. If the multi-label problem is actually a ranking problem, then this feature selection method is appropriate.
  • the mean of features is given with its standard deviation in parenthesis.
  • the number of enors counts the number of points that have been assigned to a wrong label set.
  • FIG. 21 shows the distribution of the mistakes using the leave-one-out estimate of the Hamming Loss for the Prostate Cancer Database using 4 labels with Feature selection.
  • FIG. 21a is a histogram of the enors for the direct approach where the value of one bars is 11.
  • FIG. 21b is a histogram of the enors for the binary approach, where the values of the bars are from left to right: 12 and 7. This comparison may seem a little biased since the function ⁇ (x) that computes the size of the label sets is known perfectly. Note, however, that for the binary approach, feature selection preprocessing provides a better leave-one-out Hamming Loss. There are also fewer mislabeled points than there are if no feature selection is used.
  • the inventive method can be applied to more complex problems when only dot- products are involved.
  • the kernel trick can be used to transform a linear model into a potentially highly non-linear one. To do so, it suffices to replace all the dot products (x,,Xy) by kfa j) where k is a kernel. Examples of such kernels
  • Il2 are the gaussian kernel exp (- x,- -x / ⁇ 2 )or the polynomial kernel
  • a fourth method for feature selection used the unbalanced conelation score (CORR Ub ) according to criterion
  • this criterion can also be used to assign ranlc to a subset of features rather than just a single feature. This can be done by computing the logical OR of the subset of features. If this new vector is designated as x, compute ⁇ x ⁇ .
  • the dataset used for testing of the CORR ub feature selection method concerns the prediction of molecular bioactivity for drug design.
  • the problem is to predict whether a given drug binds to a target site on thrombin, a key receptor in blood clotting.
  • the dataset was provided by DuPont
  • the data was split into a training and a test set.
  • observation has a fixed length vector of 139,351 binary features (variables) in ⁇ 0,1 ⁇ . Examples that bind are refened to as having label +1 (and hence being called positive examples). Conversely, negative examples (that do not bind) are labeled -1.
  • the test set has 634 examples.
  • the task is to determine which of the features are critical for binding affinity and to accurately predict the class values using these features.
  • Performance is evaluated according to a weighted accuracy criterion due to the unbalanced nature of the number of positive and negative examples. That is, the score of an estimate Y of the labels Y is:
  • the base classifier is a SVM.
  • K (X *X'+l) d
  • X is the matrix of fraining data.
  • methods are used to control fraining enor.
  • To introduce decision rules which allow some fraining enor (SVMs otherwise try to conectly separate all the data) one can modify the kernel matrix by adding a diagonal term. This controls the fraining enor in the following way.
  • An SVM mn imizes a regularization tenn R (a term which measures the smoothness of the chosen function) and a training error term multiplied by a constant C : R + C * L.
  • ROCS receiver operating characteristics
  • Table 11 Note that the test score is particularly low on the 2Q- C method. Recall that the balance between classifying negative and positive points correctly is controlled by the diagonal term added to the kernel, which was fixed a priori. It is possible that by controlling this hyperparameter one could obtain better results, and it may be that the different systems are unfairly reflected, however this requires retraining with many hyperparameters. Compensation for the lack of tuning was attempted by controlling the threshold of the real valued function after training. In this way one obtains control of the number of false negatives and false positives so that as long as the classifier has chosen roughly the correct direction (the conect features and resulting hyperplane), results can improve in terms of the success rate. To verify this method, this parameter was adjusted for each of the algorithms, then the maximum value of the weighted CV (cross validation) success rate was talcen. This is shown in Table 12 with the values cv max , train raax and test max .
  • Table 15c In Tables 16a-c , the same results are provided showing the maximum value of the weighted success rate with respect to changing the constant factor added to the real valued output before thresholding the decision rule (to control the tradeoff between false positives and false negatives).
  • the best success rate found on the test set using this method is 75.84%.
  • Cross-validation (CV) results indicate that 9 or 10 features produce the best results, which gives a test score of 74.49%.
  • the training score does not improve as expected when more complicated models are chosen. This is likely the result of two factors: first, the size of the diagonal term on the kernel may not scale the same resulting in a different value of the cost function for training enors, and second (which is probably the more important reason) fraining error is only approximately minimized via the cost function employed in SVM which may suffer particularly in this case of unbalanced fraining data.
  • the distance measure was altered so that if x t is a positive example, the measure of distance to x was scaled by a parameter ⁇ .
  • ⁇ By controlling ⁇ , one controls the importance of the positive class.
  • the value of ⁇ could be found by maximizing over success rates on the fraining set, but in the present experiments, the maximum performance on the test set over d e possible choices of ⁇ to get an upper bound on the highest attainable success rate of k- NN was observed.
  • the results for this method (k-NN max ) and conventional /c-NN are given in Table 17. These results fall short of the previously-described SVM results (0.7449) with SVM outperforming both variations of &-NN for all values of the hyperparameters. k 1 2 3 4 5 6 7 8
  • the present embodiment of the feature selection method was compared against another standard: correlation coefficients. This method ranks features according to the score
  • the thresholds t ⁇ and t 2 should be chosen according to the conditional probability of y given the distance from the hyperplane.
  • the thresholds were hand-selected based upon the distribution of correctly labeled examples in cross validation experiments. Feature selection is performed using CORR U b using both the training set and the sets A and B to provide a sort of enlarged training set.
  • a SVM is trained on the original training set with the n best features from the resulting correlation scores. For each n, the best scores of the SVM were calculated by adjusting the threshold parameter.
  • FIG 23 shows the results of the transductive CORR U b 2 method compared to the inductive version for from 4 to 100 features.
  • the best success rate for the transductive algorithm is 82.86%).
  • the transductive results provide improvement over the inductive method.
  • the transductive results are particularly robust with increasing numbers of features, up to one hundred features. After 200 features the results start to diminish, yielding only 77% success.
  • the inductive methods give 50%> success, i.e., they no longer learn any structure. For 1000 features using the transductive method one obtains 58%) and for 10000 features one no longer learns, obtaining 50%.
  • feature selection scores and forward selection schemes which talce into account the unbalanced nature of the data, as in the present example, are less likely to overfit than more complex backward selection schemes. Transduction is useful in talcing into account the different distributions of the test sets, helping to select the correct features.
  • the present method of feature selection comprises means for estimating the number of variables falsely called "significant" when variable selection is performed according to the SF-SVM (single feature support vector machine) criterion, h the case of the specific data, the goal is to select a subset of variables from a large costly microanay to construct less costly microanays and perform more extensive experiments.
  • the variables are ranked in order of best fit to achieve this separation and predict the fraction of false positives. While the present method is described in terms of gene selection, it should be noted that the method may be similarly applied to selection of other features.
  • a high cost array with a larger number of variables n '»n is then chosen.
  • the number p of experiments to be run on the large cost arrays is determined.
  • the p experiments are run on the n ' variable anays and the SF-SNM method to rank the variables is used.
  • the top n most promising variables are selected.
  • the fraction g of falsely significant genes using the method outlined in the report are estimated. If g>f more experiments on the large costly arrays need to be run and the whole procedure is iterated until g ⁇ f
  • Example 8 Renal cancer dataset
  • the present example selects a subset of genes that are most characteristic of renal malignancies.
  • the data consists of gene expression coefficients recorded with an expensive DNA microanay technology for a very small number of tumors (7 samples).
  • the goal is to select the most promising subset of genes to design less expensive microanays which include the selected subset of genes only, in order to conduct more extensive experiments. Results are presented using
  • Positive numbers mean upregulation of the gene in cancer compared to normal, negative numbers indicate downregulation of the gene in cancer compared to normal (Ratios have been log 2 transformed).
  • the data includes 5312 gene expression coefficients per tumor along with clone ID and a brief gene description for each coefficient is also listed.
  • Tumors are traditionally grouped according to cytogenic abnormalities.
  • Oncocytoma two samples, types A and B; and Chromophobe (1 sample). The characteristics of the types of cells are summarized below:
  • the Robson staging classification divides stages into: confinement to the renal parenchyma (stage I), tumor extension into the perirenal fat (stage IT), tumor involvement of the renal vein or inferior vena cava (stage Hla), or tumor involvement of local hilar lymph node or other vascular structures (stage Bib).
  • Stage rv classifies tumors involving adjacent organs or distant metastasis.
  • RCC develops in approximately one-third of patients with von Hippel-Lindau Syndrome. This is a deletion in the 3p region of the chromosome where a tumor suppressor gene is located. Mutations or deletions here can cause a loss of control, allowing certain cancers to develop where they would ordinarily be suppressed.
  • Oncocytoma This type represents 5% of tumors which are low grade tumors which they have malignant potential.
  • the problem can be viewed as a two class or multiclass problem.
  • Conventional cells always stand alone as a distinct entity.
  • the chromophobe and oncocytomas can be grouped together because they are much more similar to each other than to the conventional type, h the first part of the study, the examples were split into two classes: all "conventional" (samples 1-4), and the others (samples 5-7).
  • Multi-class methods were used to analyze the second phase data set collected using a smaller subset of pre-selected genes but a larger number of tumors.
  • the present method of feature selection provides for ranlcing of genes according to how well they separate tumor categories, making direct use of the knowledge of these categories in a supervised learning scheme.
  • the SF-SVM uses the size of the gap (margin) between classes as a ranking criterion. It is compared with a reference method using differences in mean class values. Both criteria use normalized gene expressions to account for each "gene-specific scatter", which is not talcen into account in the fold change criterion. The reliability of the ranlcing is then quantified statistically.
  • the data was pre-processed by successively normalizing the columns, the rows and then, again, the columns of the matrix. Normalization consists of subtracting the mean and dividing by the standard deviation.
  • the seven tumor examples were plotted in the space of the first two gene eigenvectors, both for non-normalized and normalized data.
  • the first class tumors 1-4
  • the second class tumors 5-7
  • the grouping of the tumors into sub-categories was not apparent.
  • Figure 24 provides the scatter plot for the normalized data for the different tissue types separated into two classes according to the first principal component (horizontal axis.)
  • the circle designates Conventional, stage II ("Conv. II")
  • the solid square designates Conv. HI
  • the plus sign conesponds to Conv. TV For class 2, the open square conesponds to oncocytoma and the open diamond corresponds to chromophobe.
  • the preceding confirms the results obtained by clustering and supports the fact that tumors 1-4 are well separated from tumors 5-7
  • class 1 (tumors 1-4) and class 2 (tumors 5-7) are easily separated. Genes are ranked according to how significantly they separate one class from the other. Using the SVM criterion, the number of genes falsely called significant in any given list of genes can be estimated. The result is compared to that obtained by the classical t-test method as a reference statistical method.
  • the size of the data set (seven tumors) does not permit classical machine learning analysis in which the data are split into a training set and a test set, or even to perform cross-validation experiments. Therefore, classical statistics methods of gene ranlcing evaluation were used. This differs from previous studies in which a novel criterion for gene ranlcing developed for prostate cancer data analysis was used.
  • Single Feature SVM ranks the genes according to their margin value, i.e., the distance between the extremal points of the two classes under study. For example, assume a gene is, on average, overexpressed for class 1 examples and underexpressed for class 2 examples. The margin is computed as the difference between the minimum value of the class 1 examples and the maximum value of the class 2 examples. To perform that analysis, it is important that the gene expression coefficients be normalized according to the pre-processing steps described above.
  • SF-SVM provides a better confidence that the genes selected are truly significant. Values for the top genes that were selected using SF-SVM are given in Tables 23 and 24. Table 23 includes genes underexpressed for class 1 (conventional) and overexpressed for class 2 (chromophobe/oncocytoma), while Table 24 lists genes overexpressed for class 1 and underexpressed for class 2. hi both tables, the "Margin” is the SF-SVM ranlcing criterion. Its exponentiated value "Expmar” is also provided. All genes listed have a p-value less than 0.001.
  • the sixth through nineteenth genes in Table 23 have a p-value less than 0.0005 and first five genes in the list have a p-value less than 0.0001.
  • all genes have a probability less than 0.001 to be false positive.
  • the first three genes listed in Table 24 have a probability of less than 0.0001 to be false positive, while the fourth through fourteenth entries have a probability of less than 0.0005 to be false positive.
  • FIG. 25 is a plot of the distribution of margin values for the combination of samples used to build Table 25. A few values extracted from this plot are shown in Table 25 For example, only 0.01 (1%) of "random" genes have a margin exceeding 0.41. A larger fraction of real genes separate the two classes with a margin exceeding that value (398 genes, that is 7% of the 5312 in the data set). Thus, among these 398 genes that have margin values exceeding 0.41, it is expected that, at most, 53 genes will be falsely called significant.
  • All methods ranlc genes according to how well the examples of one class are separated from the examples of the other class using a given criterion.
  • the genes are indexed with the letter i and w,- is the ranlcing criterion for a given gene.
  • (+) denotes the class for which gene i is overexpressed on average and class and (-) the other class ⁇ ; is denoted by the sign of w, (gene polarity), which is +1 if class (+) coincides with class 1 and -1 otherwise.
  • Genes with positive W t are overexpressed for class 1 and underexpressed for class 2.
  • Two types of criteria are compared, which are the difference in mean values and the difference in extremal values.
  • Difference in mean values Several methods are based on the difference in mean expression value of the two classes ⁇ i(+) and ⁇ i(-), normalized with a coefficient that reflects the intrinsic scatter of gene expression values of that gene. Typically, an average of the two infra-class standard deviations ⁇ i(+) and ⁇ i(-) is considered, e.g.
  • Sj(+) is defined as the smallest observed value of class (+) and s;(-) the largest observed value of class (-).
  • FIG. 26 provides plots of the hypothetical distributions of gene expression coefficients of a given gene for two categories of samples, where class (-) is indicated by the "•” on the curve and class (+) by the "+” on the curve. Sets of examples drawn from those distributions are indicated by "•” and "+” on the horizontal anow. Criteria to determine whether a given gene separates the two classes are derived by examining either the difference in mean values ⁇ i(+) - ⁇ i(-) or the difference in extremal values S ⁇ (+) - s ⁇ (-).
  • FIG 26a illustrates a typical sample drawn from well separated classes.
  • FIG. 26b models a purely insignificant gene by drawing at random examples of both classes from the same distribution N(0,1). It is unlikely that the means of the examples of the two classes will be well separated. It is even more unlikely that the extremal values will be separated by a positive margin.
  • a gene is called "significant" for the class separation at hand if its criterion exceeds a certain threshold value.
  • the estimated p- value is the fraction of drawings in which the criterion exceeds a given threshold. Therefore, for any criterion threshold value, it is possible to obtain an upper bound estimate of the number of genes called significant that are plotted as a function of the number of genes called significant.
  • FIG. 27 provides a plot of an estimated upper bound on the number of genes falsely called significant as a function of the number of genes called significant.
  • the estimate uses 100000 genes drawn at random according to N(0,1).
  • the results using the SF-LDA (t- test) method are indicated by the curve with circles.
  • the other line indicates the results of the SF-SVM, with the dashed portion of that line, below the intersection of the two lines, conesponding to genes that separate the two classes perfectly.
  • the SF-SVM criterion that uses the difference between extremal points as ranlcing criterion incurs a smaller number of genes falsely called significant than does the SF-LDA criterion that uses the difference of the means. Above the point where the two curves cross, SF-LDA becomes more reliable.
  • the SF-SVM criterion appears to be slightly superior to the SF-LDA criterion becauase, for a given number of genes called significant, it provides a smaller estimated number of genes falsely called significant where positive margins (genes separating perfectly the two classes) can be defined.
  • genes can be ranked according a criterion that characterize how well they individually discriminate between classes.
  • two types of methods can be used: (1) methods based on differences in mean values; and (2) methods based on margins
  • the well known Fisher criterion or Linear Discriminant Analysis (LDA) criterion pertains to the first method, while the SVM criterion pertain to the second method.
  • the multi-class criteria are generalizations of the two-class criteria explained previously. In the experiments, the data is first nonnalized as described above. The problem is considered as having five classes (in this case, diseases) based on the five tissue types discussed above: Conventional fl " (2 patients); Conventional m
  • T-cell differentiation protein ⁇ Incyte PD:504786 ⁇ lectin, galactoside-binding, soluble, 3 (galectin 3) ⁇ Incyte PD:2921194 ⁇ abundant in neuroepithelium area ⁇ Incyte PD:637576 ⁇
  • FIG. 28a shows the results for small inducible cytokine A2 (monocyte chemotactic protein 1, homologous to mouse Sig-je), ⁇ fricyte PD: 1511342 ⁇ .
  • FIG. 28b shows the results for ATP synthase, H+ transporting, mitochondrial Fl complex, alpha subunit, isoform 1, cardiac muscle ⁇ Incyte PD:3206210 ⁇ .
  • the present method of feature selection uses statistical methods to estimate the fraction of genes that might be falsely called significant in a given list of genes that appear to separate well the 4 examples of conventional RCC from the 3 examples of chromophobe or oncocytoma.
  • the two methods used are the conventional t-test (SF-LDA) and SF-SVM. Both methods are in good agreement, however, genes that separate perfectly the two classes identified by SF-SVM do so with a smaller predicted number of genes falsely called significant.
  • the gene ranlcing provided can be used to select genes that are most promising to build a new microanay with a more directed approach. It should be noted that, because of the small number of examples available, there is a certain degree of uncertainty as to the validity of the genes identified.
  • a number of different methods are provided for selection of features for use in a learning machine using data that best represents the essential infonnation to be exfracted from the data set.
  • the inventive methods provide advantages over prior art feature selection methods by talcing into account the intenelatedness of the data, e.g., multi-label problems.
  • the features selection can be performed as a pre-processing step, prior to training the learning machine, and can be done in either input space or feature space.

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Software Systems (AREA)
  • Data Mining & Analysis (AREA)
  • Evolutionary Computation (AREA)
  • General Physics & Mathematics (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Physics & Mathematics (AREA)
  • Artificial Intelligence (AREA)
  • General Engineering & Computer Science (AREA)
  • Mathematical Physics (AREA)
  • Medical Informatics (AREA)
  • Computing Systems (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Evolutionary Biology (AREA)
  • Investigating Or Analysing Biological Materials (AREA)

Abstract

Des caractéristiques sont prétraitées (204) de manière à réduire les erreurs de classification dans des machines à vecteur de support (SVM) (200) utilisées pour identifier des formes dans des grandes bases de données. Le prétraitement (204) est exécuter de manière obliger les caractéristiques utilisées à former (210) la machine d'apprentissage à vecteur de support. Des données réelles (226) sont collectées et traitées (232) au moyen de la machine à vecteur de support.
EP02778747A 2001-11-07 2002-11-07 Classement de caracteristiques pretraitees pour une machine a vecteur de support Withdrawn EP1449108A4 (fr)

Applications Claiming Priority (3)

Application Number Priority Date Filing Date Title
US34756201P 2001-11-07 2001-11-07
US347562P 2001-11-07
PCT/US2002/035576 WO2003040949A1 (fr) 2001-11-07 2002-11-07 Classement de caracteristiques pretraitees pour une machine a vecteur de support

Publications (2)

Publication Number Publication Date
EP1449108A1 EP1449108A1 (fr) 2004-08-25
EP1449108A4 true EP1449108A4 (fr) 2006-11-22

Family

ID=23364249

Family Applications (1)

Application Number Title Priority Date Filing Date
EP02778747A Withdrawn EP1449108A4 (fr) 2001-11-07 2002-11-07 Classement de caracteristiques pretraitees pour une machine a vecteur de support

Country Status (2)

Country Link
EP (1) EP1449108A4 (fr)
WO (1) WO2003040949A1 (fr)

Families Citing this family (11)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2005048185A1 (fr) * 2003-11-17 2005-05-26 Auckland University Of Technology Methode d'inference neuro-floue transductive pour la modelisation personnalisee
NZ572036A (en) 2008-10-15 2010-03-26 Nikola Kirilov Kasabov Data analysis and predictive systems and related methodologies
BRPI1015129A2 (pt) 2009-06-30 2016-07-12 Dow Agrosciences Llc aplicação de métodos em aprendizagem de máquina para regras de associação na mineração de conjuntos de dados contendo marcadores genéticos moleculares de plantas e de animais, seguida pela classificação ou predição utilizando atributos criados a partir destas regras de associação
US10535014B2 (en) 2014-03-10 2020-01-14 California Institute Of Technology Alternative training distribution data in machine learning
US9953271B2 (en) 2013-11-22 2018-04-24 California Institute Of Technology Generation of weights in machine learning
US10558935B2 (en) 2013-11-22 2020-02-11 California Institute Of Technology Weight benefit evaluator for training data
US9858534B2 (en) 2013-11-22 2018-01-02 California Institute Of Technology Weight generation in machine learning
US9658987B2 (en) 2014-05-15 2017-05-23 International Business Machines Corporation Regression using M-estimators and polynomial kernel support vector machines and principal component regression
US11139048B2 (en) * 2017-07-18 2021-10-05 Analytics For Life Inc. Discovering novel features to use in machine learning techniques, such as machine learning techniques for diagnosing medical conditions
CN111329847A (zh) * 2020-03-19 2020-06-26 上海大学 利用二氢查尔酮类化合物对胰岛素促泌性能进行预报的方法及应用
CN111652393A (zh) * 2020-06-05 2020-09-11 国网信通亿力科技有限责任公司 基于大数据技术的电力设备异常预警方法

Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2001005935A2 (fr) * 1999-07-16 2001-01-25 Rosetta Inpharmatics, Inc. Conception de sonde iterative et etablissement de profils d'expression detailles avec jeux ordonnes d'echantillons adaptables de synthese in-situ
WO2001031579A2 (fr) * 1999-10-27 2001-05-03 Barnhill Technologies, Llc Procedes et dispositifs permettant d'identifier des motifs dans des systemes biologiques et procedes d'utilisation correspondants

Family Cites Families (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US6192360B1 (en) * 1998-06-23 2001-02-20 Microsoft Corporation Methods and apparatus for classifying text and for building a text classifier

Patent Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2001005935A2 (fr) * 1999-07-16 2001-01-25 Rosetta Inpharmatics, Inc. Conception de sonde iterative et etablissement de profils d'expression detailles avec jeux ordonnes d'echantillons adaptables de synthese in-situ
WO2001031579A2 (fr) * 1999-10-27 2001-05-03 Barnhill Technologies, Llc Procedes et dispositifs permettant d'identifier des motifs dans des systemes biologiques et procedes d'utilisation correspondants

Non-Patent Citations (3)

* Cited by examiner, † Cited by third party
Title
EISEN M: "Cluster and TreeView Manual", 1998 - 1999, pages 1 - 20, XP002402140, Retrieved from the Internet <URL:http://rana.lbl.gov/manuals/ClusterTreeView.pdf> [retrieved on 20061006] *
See also references of WO03040949A1 *
YOUNG A N ET AL: "Expression profiling of renal epithelial neoplasms/A METHOD FOR TUMOR CLASSIFICATION AND DISCOVERY OF DIAGNOSTIC MOLECULAR MARKERS", AMERICAN JOURNAL OF PATHOLOGY, PHILADELPHIA, PA, US, vol. 158, no. 5, May 2001 (2001-05-01), pages 1639 - 1651, XP002962291, ISSN: 0002-9440 *

Also Published As

Publication number Publication date
EP1449108A1 (fr) 2004-08-25
WO2003040949A1 (fr) 2003-05-15

Similar Documents

Publication Publication Date Title
US7624074B2 (en) Methods for feature selection in a learning machine
US7318051B2 (en) Methods for feature selection in a learning machine
US7805388B2 (en) Method for feature selection in a support vector machine using feature ranking
US7475048B2 (en) Pre-processed feature ranking for a support vector machine
US8095483B2 (en) Support vector machine—recursive feature elimination (SVM-RFE)
Shmilovici Support vector machines
JP5064625B2 (ja) パターンを同定するための方法及び機械
US8463718B2 (en) Support vector machine-based method for analysis of spectral data
US7617163B2 (en) Kernels and kernel methods for spectral data
US6658395B1 (en) Enhancing knowledge discovery from multiple data sets using multiple support vector machines
EP1192595B1 (fr) Amelioration de la decouverte de connaissances a partir d&#39;ensembles de donnees multiples au moyen de machines a vecteurs de soutien multiples
WO2002091211A1 (fr) Noyaux et procedes de selection de noyaux a utiliser dans des machines a enseigner
Osareh et al. Microarray data analysis for cancer classification
WO2001031579A2 (fr) Procedes et dispositifs permettant d&#39;identifier des motifs dans des systemes biologiques et procedes d&#39;utilisation correspondants
CA2435254C (fr) Procedes d&#39;identification de motifs dans des systemes biologiques et utilisations desdits procedes
Douzas et al. Geometric SMOTE: Effective oversampling for imbalanced learning through a geometric extension of SMOTE
EP1449108A1 (fr) Classement de caracteristiques pretraitees pour une machine a vecteur de support
AU2002253879A1 (en) Methods of identifying patterns in biological systems and uses thereof
AU3783099A (en) Pre-processing and post-processing for enhancing knowledge discovery using support vector machines
Altınçay Decision trees using model ensemble-based nodes
Pérez-Sánchez et al. Selecting target concept in one-class classification for handling class imbalance problem
Hassan Regularization in Machine Learning with Applications in Biology
Renukadevi et al. INVESTIGATING THE PERFORMANCE OF OPTIMIZATION TECHNIQUES WITH SVM FOR MEDICAL IMAGE CLASSIFICATION

Legal Events

Date Code Title Description
PUAI Public reference made under article 153(3) epc to a published international application that has entered the european phase

Free format text: ORIGINAL CODE: 0009012

17P Request for examination filed

Effective date: 20040517

AK Designated contracting states

Kind code of ref document: A1

Designated state(s): AT BE BG CH CY CZ DE DK EE ES FI FR GB GR IE IT LI LU MC NL PT SE SK TR

AX Request for extension of the european patent

Extension state: AL LT LV MK RO SI

RIN1 Information on inventor provided before grant (corrected)

Inventor name: GUYON, ISABELLE

Inventor name: PEREZ-CRUZ, FERNANDO,A.T.S.C.,UNIV. CAROLOS III

Inventor name: SCHOELKOPF, BERNHARD

Inventor name: ELLISSEEF, ANDRE

Inventor name: WESTON, JASON

RAP1 Party data changed (applicant data changed or rights of an application transferred)

Owner name: BIOWULF TECHNOLOGIES, LLC

RAP1 Party data changed (applicant data changed or rights of an application transferred)

Owner name: MCKENZIE, JOE

Owner name: CARLS, GARRY L.

Owner name: BERGERON, GLYNN

Owner name: O'HAYER, TIMOTHY P.

Owner name: SIMPSON, K. RUSSELL

Owner name: MATTHEWS, JOHN E.

Owner name: ANDERSON, CURTIS

Owner name: FARLEY, PETER J.

Owner name: PADEREWSKI, JULES B.

Owner name: ROBERTS, JAMES

Owner name: STERN, JULIAN N.

Owner name: MEMORIAL HEALTH TRUST, INC.

RAP1 Party data changed (applicant data changed or rights of an application transferred)

Owner name: HEALTH DISCOVERY CORPORATION

RIC1 Information provided on ipc code assigned before grant

Ipc: G06F 15/18 20060101ALI20061010BHEP

Ipc: G06F 19/00 20060101AFI20061010BHEP

A4 Supplementary search report drawn up and despatched

Effective date: 20061019

17Q First examination report despatched

Effective date: 20070410

RAP1 Party data changed (applicant data changed or rights of an application transferred)

Owner name: HEALTH DISCOVERY CORPORATION

STAA Information on the status of an ep patent application or granted ep patent

Free format text: STATUS: THE APPLICATION IS DEEMED TO BE WITHDRAWN

18D Application deemed to be withdrawn

Effective date: 20110826