CN111340240A - Method and device for realizing automatic machine learning - Google Patents
Method and device for realizing automatic machine learning Download PDFInfo
- Publication number
- CN111340240A CN111340240A CN202010219796.5A CN202010219796A CN111340240A CN 111340240 A CN111340240 A CN 111340240A CN 202010219796 A CN202010219796 A CN 202010219796A CN 111340240 A CN111340240 A CN 111340240A
- Authority
- CN
- China
- Prior art keywords
- configuration information
- machine learning
- user
- parameter
- parameter configuration
- 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.)
- Pending
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
Abstract
A method and apparatus for implementing automatic machine learning is disclosed. Acquiring process configuration information set by a user and used for characterizing at least part of a machine learning modeling process, wherein the process configuration information comprises one or more operation steps; acquiring parameter configuration information set by a user aiming at the at least part of machine learning modeling process, wherein the parameter configuration information comprises a value space of at least part of operation parameters related to the operation step; executing the at least part of the machine learning modeling process based on the process configuration information and different value combinations of the at least part of the operating parameters determined based on the parameter configuration information to obtain a plurality of execution results; and outputting the preferred values of at least part of the operating parameters according to the execution results. Therefore, automatic machine learning can be realized by acquiring the process configuration information and the parameter configuration information set by the user, and the development cost of the automatic machine learning can be reduced.
Description
Technical Field
The present invention relates generally to the field of artificial intelligence, and more particularly, to a method and apparatus for implementing automatic machine learning.
Background
Machine learning is a necessary product of the development of artificial intelligence research to a certain stage, and aims to improve the performance of the system by means of calculation and by using experience. In a computer system, "experience" is usually in the form of "data" from which a "model" can be generated by a machine learning algorithm, i.e. by providing empirical data to a machine learning algorithm, a model can be generated based on these empirical data, which provides a corresponding judgment, i.e. a prediction, in the face of a new situation.
A machine learning model generally solves the problem in a specific scenario, and development of the machine learning model requires a large investment in human resources and a special human resource. In response to the deficiencies of conventional machine learning modeling schemes, automatic machine learning (AutoML) has been developed, the objective of which is to determine machine learning solutions using automated data-driven approaches.
However, the development cost of the automatic machine learning is high at present, and how to enable a user to realize the automatic machine learning through simple operation so as to reduce the development cost of the automatic machine learning is a problem to be solved urgently at present.
Disclosure of Invention
Exemplary embodiments of the present invention are directed to providing a scheme for implementing automatic machine learning in which a machine learning model is online, so as to reduce the cost of developing automatic machine learning for a user.
According to a first aspect of the present invention, a method of implementing automatic machine learning is presented, comprising: acquiring process configuration information set by a user and used for representing at least part of a machine learning modeling process, wherein the process configuration information comprises one or more operation steps; acquiring parameter configuration information set by a user aiming at least part of a machine learning modeling process, wherein the parameter configuration information comprises a value space of at least part of operation parameters related to an operation step; executing at least part of the machine learning modeling process based on the process configuration information and different value combinations of at least part of the operation parameters determined based on the parameter configuration information to obtain a plurality of execution results; and outputting the preferred values of at least part of the operating parameters according to the execution results.
Optionally, the step of obtaining user-set process configuration information characterizing at least part of the machine learning modeling process comprises: providing a data uploading interface for a user; receiving a file uploaded by a user through a data uploading interface, wherein the file is written by the user for at least part of the machine learning modeling process based on a language of a specific rule; and analyzing the file to determine the process configuration information.
Optionally, the file further defines a value space of at least part of the operating parameters involved in the operation step, and the step of obtaining the parameter configuration information set by the user for at least part of the machine learning modeling process includes: and analyzing the file to determine parameter configuration information.
Optionally, the step of obtaining user-set process configuration information characterizing at least part of the machine learning modeling process comprises: presenting an interactive interface for setting up at least part of a machine learning modeling process to a user; and acquiring process configuration information set by a user through an interactive interface.
Optionally, the step of obtaining parameter configuration information set by the user for at least part of the machine learning modeling process includes: displaying an interactive interface for setting a value space of an operation parameter related to the operation step to a user; and acquiring parameter configuration information set by a user through an interactive interface.
Optionally, the process configuration information further defines at least one target node, the target node being a data node, wherein the step of performing at least part of the machine learning modeling process includes iteratively performing the following operations based on different combinations of values of at least part of the operating parameters determined by the process configuration information and based on the parameter configuration information: determining one or more groups of value combinations of at least part of the operating parameters according to the parameter configuration information; and executing at least part of the machine learning modeling process according to the process configuration information and the determined value combination to obtain an execution result of the target node.
Optionally, in each iteration, the method further comprises: the steps of adding the value combination and the execution result determined in the current round to the historical information, and determining one or more groups of value combinations of at least part of the operation parameters according to the parameter configuration information comprise: and determining one or more groups of value combinations of at least part of the operating parameters of the current round by using a preset strategy according to the parameter configuration information and the historical information.
Optionally, the policy comprises at least one of: a random strategy; an enumeration strategy; a Bayesian optimization strategy; and (4) grid strategy.
Optionally, the method further comprises: saving the execution result of the operation step with unchanged value of the operation parameter; and in the process of executing at least part of machine learning modeling, ignoring the execution of the operation step of which the value of the operation parameter is unchanged in the process of executing at least part of machine learning modeling, and calling the execution result of the operation step to be stored in advance.
Optionally, the method further comprises: estimating the resource use condition of the computing resource distributed for the operation step according to the execution result of at least part of the machine learning modeling process; and adjusting the computing resources distributed for the operation steps according to the estimated result.
Optionally, the method further comprises: obtaining screening condition configuration information set by a user and used for screening values of at least part of operating parameters, wherein the step of executing at least part of machine learning modeling process comprises the following steps: screening different value combinations of at least part of the operation parameters determined based on the parameter configuration information according to the screening condition configuration information; and executing at least part of the machine learning modeling process based on the process configuration information and the value combination after screening.
Optionally, the operation step includes an operation node and a data node, where the input and the output of the operation node are both data nodes, and the parameter configuration information includes a value space of at least part of operation parameters related to the operation node.
Optionally, the process configuration information is a computational graph composed of data nodes and operation nodes.
According to a second aspect of the present invention, there is provided an apparatus for implementing automatic machine learning, comprising: the system comprises a first acquisition module, a second acquisition module and a third acquisition module, wherein the first acquisition module is used for acquiring process configuration information which is set by a user and used for characterizing at least part of a machine learning modeling process, and the process configuration information comprises one or more operation steps; the second acquisition module is used for acquiring parameter configuration information set by a user aiming at least part of the machine learning modeling process, and the parameter configuration information comprises a value space of at least part of operation parameters involved in the operation steps; the execution module is used for executing at least part of the machine learning modeling process based on the process configuration information and different value combinations of at least part of the operation parameters determined based on the parameter configuration information to obtain a plurality of execution results; and the output module is used for outputting the optimal values of at least part of the operating parameters according to a plurality of execution results.
Optionally, the apparatus further comprises: the system comprises a providing module, a data uploading module and a data uploading module, wherein the providing module is used for providing a data uploading interface for a user; the receiving module is used for receiving a file uploaded by a user through the data uploading interface, wherein the file is written by the user aiming at least part of the machine learning modeling process based on a language of a specific rule, and the first obtaining module analyzes the file and determines process configuration information.
Optionally, the file further defines a value space of at least part of the operating parameters involved in the operating step, and the second obtaining module analyzes the file to determine the parameter configuration information.
Optionally, the apparatus further comprises: the first display module is used for displaying an interactive interface for setting at least part of the machine learning modeling process to a user, wherein the first acquisition module acquires process configuration information set by the user through the interactive interface.
Optionally, the apparatus further comprises: and the second display module is used for displaying an interactive interface for setting the value space of the operation parameters related to the operation steps to the user, wherein the second acquisition module acquires the parameter configuration information set by the user through the interactive interface.
Optionally, the process configuration information further defines at least one target node, where the target node is a data node, and the execution module iteratively executes the following operations: determining one or more groups of value combinations of at least part of the operating parameters according to the parameter configuration information; and executing at least part of the machine learning modeling process according to the process configuration information and the determined value combination to obtain an execution result of the target node.
Optionally, in each iteration process, the execution module adds the value combinations and execution results determined in the current round to the historical information, and determines one or more value combinations of at least part of the operating parameters in the current round by using a predetermined strategy according to the parameter configuration information and the historical information.
Optionally, the policy comprises at least one of: a random strategy; an enumeration strategy; a Bayesian optimization strategy; and (4) grid strategy.
Optionally, the apparatus further comprises: and the storage module is used for storing the execution result of the operation step with unchanged values of the operation parameters, and the execution module ignores the execution of the operation step with unchanged values of the operation parameters in at least part of the machine learning modeling process and calls the execution result of the operation step which is stored in advance.
Optionally, the apparatus further comprises: the estimation module is used for estimating the resource use condition of the computing resources distributed for the operation steps according to the execution result of at least part of the machine learning modeling process; and the adjusting module is used for adjusting the computing resources distributed to the operation steps according to the estimation result.
Optionally, the apparatus further comprises: the third acquisition module is used for acquiring screening condition configuration information which is set by a user and used for screening values of at least part of operating parameters; the screening module is used for screening different value combinations of at least part of the operation parameters determined based on the parameter configuration information according to the screening condition configuration information; the execution module executes at least part of the machine learning modeling process based on the process configuration information and the screened value combination.
Optionally, the operation step includes an operation node and a data node, where the input and the output of the operation node are both data nodes, and the parameter configuration information includes a value space of at least part of operation parameters related to the operation node.
Optionally, the process configuration information is a computational graph composed of data nodes and operation nodes.
According to a third aspect of the present invention, there is provided a system comprising at least one computing device and at least one storage device storing instructions, wherein the instructions, when executed by the at least one computing device, cause the at least one computing device to perform the method as set forth in the first aspect of the present invention.
According to a fourth aspect of the present invention, a computer-readable storage medium storing instructions is proposed, wherein the instructions, when executed by at least one computing device, cause the at least one computing device to perform the method according to the first aspect of the present invention.
According to the method and the device for realizing automatic machine learning, the automatic machine learning can be realized by acquiring the process configuration information and the parameter configuration information set by the user, so that the development cost of the automatic machine learning can be reduced.
Drawings
These and/or other aspects and advantages of the present invention will become more apparent and more readily appreciated from the following detailed description of the embodiments of the invention, taken in conjunction with the accompanying drawings of which:
FIG. 1 shows a flow diagram of a method of implementing automatic machine learning, according to an example embodiment of the present invention;
FIG. 2 illustrates a computational graph for characterizing a machine-learned feature selection process;
fig. 3 illustrates a block diagram of an apparatus for implementing automatic machine learning according to an exemplary embodiment of the present invention.
Detailed Description
In order that those skilled in the art will better understand the present invention, exemplary embodiments thereof will be described in further detail below with reference to the accompanying drawings and detailed description.
Fig. 1 shows a flowchart of a method of implementing automatic machine learning according to an exemplary embodiment of the present invention. The method shown in fig. 1 may be implemented entirely in software via a computer program, and the method shown in fig. 1 may also be executed by a specifically-configured computing device.
Referring to fig. 1, at step S110, process configuration information set by a user to characterize at least a portion of a machine learning modeling process is obtained, the process configuration information including one or more operational steps. Wherein different operation steps may have a predetermined execution order therebetween.
At least part of the machine learning modeling process can be a complete machine learning modeling process or a partial flow in the machine learning modeling process. By way of example, at least a portion of the machine learning modeling process may include, but is not limited to, one or more of a data stitching process, a data splitting process, a feature generation process, a model building process, a model training process, a model testing process, a model evaluation process, a model application process, and the like.
The invention can acquire the process configuration information set by the user in various ways. The invention can provide a plurality of modes for setting the process configuration information for the user so as to acquire the process configuration information set by the user in a plurality of modes. Thus, process configuration information may refer to multiple forms of information.
For example, the process configuration information may be written by a user in a language based on particular rules, i.e., the user may write the process configuration information by writing program code. As an example, a data upload interface may be provided to a user, a file uploaded by the user through the data upload interface may be obtained, and the process configuration information may be determined by parsing the file, where the file may be written in a language based on a specific rule by the user, for example, the file may refer to program code written by the user for the at least part of the machine learning modeling process, where the program code defines one or more operation steps and an execution order among the different operation steps.
For another example, the process configuration information may also be set by the user through a visual operation (e.g., dragging). By way of example, an interactive interface for setting the at least part of the machine learning modeling process may be presented to a user, and process configuration information set by the user through the interactive interface may be obtained. For example, the user may set the process configuration information by building a flowchart for characterizing at least a part of the machine information modeling process in the interactive interface, that is, the process configuration information may be the flowchart built by the user in the interactive interface. Optionally, a plurality of nodes (operation nodes and data nodes) that can be selected by a user may be displayed in the interactive interface, and the user may configure the attribute information of the nodes by selecting the nodes and configure the execution logic between the nodes in a wired manner to set the process configuration information. The specific implementation process of the content that can be displayed in the interactive interface and the process configuration information set by the user through the interactive interface is not the focus of the present invention, and therefore, the detailed description is not provided.
In step S120, parameter configuration information set by a user for at least part of the machine learning modeling process is obtained, where the parameter configuration information includes a value space of at least part of the operating parameters involved in the operating step.
The sequence of the steps S110 and S120 is not limited in the present invention. That is, step S110 may be performed first, and then step S120 may be performed; step S120 may be performed first, and then step S110 may be performed; or step S110 and step S120 may not be executed sequentially and simultaneously.
The parameter configuration information may be set by the user when the process configuration information is set, or may be set by the user after the process configuration information is set. Optionally, in the process of setting the process configuration information, the user may configure a value space of at least a part of the operation parameters related to the operation steps in the process configuration information, so as to set the parameter configuration information.
The invention can acquire the parameter configuration information set by the user in various ways. The invention can provide a plurality of modes for setting the parameter configuration information for the user so as to acquire the parameter configuration information set by the user in a plurality of modes. Thus, parameter configuration information may refer to multiple forms of information.
For example, the parameter configuration information may be written by a user in a language based on specific rules, i.e., the user may write the parameter configuration information by writing program code. As an example, a data uploading interface may be provided for a user, a file uploaded by the user through the data uploading interface may be obtained, and the parameter configuration information may be determined by parsing the file, where the file may be written in a language based on a specific rule by the user, for example, the file may refer to program code written by the user for process configuration information, where the program code defines a value space of at least a part of the operation parameters involved in one or more operation steps.
Alternatively, the file mentioned here and the file mentioned above in connection with step S110 may refer to the same file, that is, the user may upload a file including both the process configuration information and the parameter configuration information through the data upload interface, and thus, the present invention may obtain the process configuration information and the parameter configuration information by analyzing the file uploaded by the user. For example, a user may define a value space for at least a portion of the operating parameters involved in an operational step while writing process configuration information characterizing at least a portion of the machine learning modeling process based on a language of particular rules.
For another example, the parameter configuration information may also be set by the user through a visual operation. As an example, an interactive interface for setting a value space of an operation parameter related to an operation step in the process configuration information may be presented to a user, and parameter configuration information set by the user through the interactive interface may be acquired. Optionally, the user may define a value space of at least a part of the operation parameters related to the operation steps in the process configuration information while setting the process configuration information through the interactive interface, so as to set the parameter configuration information.
As an example, the operation steps in the process configuration information may include an operation node and a data node, where an input and an output of the operation node are both data nodes, and the parameter configuration information includes a value space of at least part of operation parameters related to the operation node. Thus, the process configuration information may be a computational graph composed of data nodes and operational nodes. The computational graph can be regarded as a directed acyclic graph composed of data nodes and operation nodes. The computational graph may clearly define the execution logic of at least part of the machine learning modeling process. Users can pack the original codes into data nodes and operation nodes through the base class or the method provided by the framework, and connect the data nodes and the operation nodes into a computational graph. In addition, the user can define the operation nodes and the data nodes in a dragging mode on the graphical interface to build the calculation graph. When a user defines an operation node in the calculation graph, a value space can be defined for an operation parameter related to the operation node, so that part or all of the operation parameters in the operation node are changed from a determined item to an undetermined item, and the value of the operation parameter needs to be selected from the value space. The value spaces of all the operating parameters in the calculation map are the total value spaces corresponding to the current calculation map, and generally, the product of all the value spaces constitutes the total search space.
In step S130, at least part of the machine learning modeling process is executed based on the process configuration information and different combinations of values of at least part of the operating parameters determined based on the parameter configuration information, so as to obtain a plurality of execution results.
Parameters with undetermined values exist in at least part of the machine learning modeling process represented by the process configuration information. Different value combinations of the parameters of the part to be determined can be determined based on the parameter configuration information. The process configuration information and each combination of values can form at least part of a machine learning modeling process that performs the logic determination. Therefore, the machine learning modeling process represented by the process configuration information can be executed for each value combination, so that a plurality of execution results corresponding to different value combinations are obtained.
As an example, the process configuration information also defines at least one target node, the target node being a data node. The target node is used as an evaluation index of the value combination, namely an optimization target. For example, a target node may refer to the result of execution of the operational logic of at least a portion of the machine learning modeling process characterized by the process configuration information. Generally speaking, the execution result corresponding to the target node can be converted into an evaluable scalar quantity, such as maximum or minimum, and the goal of the present invention is to expect the result of the target node to be optimal. Thus, in the process of executing step S130, one or more sets of value combinations of at least some of the operating parameters may be determined according to the parameter configuration information; and then, executing at least part of the machine learning modeling process according to the process configuration information and the determined value combination to obtain an execution result of the target node. And evaluating the quality of the currently adopted value combination according to the execution result of the target node.
The invention can calculate the value combination to be executed in the next group or array according to the value combination determined by history, and can carry out certain pruning optimization according to the existing algorithm in the process so as to reduce the size of the search space and accelerate the acquisition of the optimal value of the operation parameter. The search space refers to a set of all possible value combinations, and can be determined according to the value space represented by the parameter configuration information and the historical value combinations.
As an example, in each iteration process, the present invention may add the value combinations and execution results determined in the current round to the history information, and then determine one or more value combinations of at least some of the operation parameters in the current round using a predetermined search strategy according to the parameter configuration information and the history information. Wherein the search policy may include, but is not limited to, at least one of: random search strategy, enumeration search strategy, Bayesian optimization search strategy and grid search strategy. The implementation principle of each search strategy can be referred to the prior theoretical knowledge, and the invention is not repeated. Optionally, when the value combination of the current round is determined, an execution result obtained by executing at least part of the machine learning modeling process based on the historical value combination may also be referred to. That is, the next group or groups of value combinations can be determined according to the historical value combinations and execution results, and in the process, certain pruning optimization can be performed according to the existing algorithm to reduce the size of the search space and accelerate the obtaining of the optimal solution.
The invention can also acquire the screening condition configuration information which is set by the user and used for screening the values of at least part of the operating parameters, and the screening condition configuration information can comprise one or more screening conditions used for screening the values of at least part of the operating parameters. After the value combinations of at least part of the operating parameters are determined based on the parameter configuration information, the value combinations can be screened according to the screening condition configuration information to eliminate the value combinations which do not meet the preset standard. The removed value combinations can be regarded as being unable to be solved and discarded, namely the removed value combinations do not participate in the execution of at least part of the machine learning modeling process any more. For example, the screening condition configuration information may include a screening condition for screening a parameter value causing a large resource consumption, where the screening condition may limit a running time consumption of the operation step and/or a range of memory usage, such as limiting the running time consumption to be lower than a first threshold and/or limiting the memory usage to be lower than a second threshold. If the running time of one or more operation steps under a certain group of value combinations is too long (for example, greater than a first threshold) or the memory occupation is too large (for example, greater than a second threshold), which exceeds the range limited by the screening condition, the group of value combinations can be discarded as an infeasible solution and not executed, that is, the final scoring result is not required to be obtained. Therefore, under the condition that the total value space of at least part of the operation parameters is large, so that the search is uncertain and high, the search process can be more controllable to a certain extent according to the screening condition configuration information set by the user, and the finally obtained optimal value is more in line with the actual requirement.
In step S140, preferred values of at least some of the operating parameters are output according to the plurality of execution results.
The execution result can be used as an evaluation index for reflecting the degree of superiority and inferiority of the adopted value combination. According to the performance of the execution result, the value combination corresponding to the execution result with the best performance or better performance is selected from the execution results to serve as the preferred value of the at least part of the operation parameters.
The machine learning modeling process characterized by process configuration information may be viewed as being comprised of a plurality of operational steps. For different value combinations, there may be operation steps with unchanged operation parameter values in the process of executing machine learning modeling, and for these operation steps, the execution result is generally fixed and unchanged. Therefore, in view of saving computing resources, the method and the device can also save the execution result of the operation step with unchanged values of the operation parameters, and can neglect the execution of the operation step with unchanged values of the operation parameters in the at least part of machine learning modeling process and call to pre-save the execution result of the operation step in the at least part of machine learning modeling process.
By way of example, the invention may also predict resource usage of the computing resources allocated for the operation steps based on the results of the execution of the at least part of the machine learning modeling process; and adjusting the computing resources distributed for the operation steps according to the estimated result. For example, in the case where the utilization rate of the computing resource allocated to a certain operation step is found to be not high after a plurality of executions of the operation step, the computing resource allocated to the operation step may be reduced.
The implementation flow of the method for implementing automatic machine learning according to the present invention is briefly described with reference to fig. 1. The method for realizing automatic machine learning can be realized as a universal parameter search framework, and the parameter search framework can be suitable for scenes such as AuotML and the like needing a large amount of parameter tuning so as to solve the problem of multi-parameter search tuning.
Taking the above-mentioned process configuration information as an example of a computation graph, the parameter search framework of the present invention can be divided into four functional modules, including a computation graph module, a parameter space module, a search policy module, and an execution engine.
The computation graph module is used for acquiring a user-defined computation graph. The computational graph is a directed acyclic graph formed by data and operations, wherein the input and output of an operation node are data nodes, and the data nodes are used as the input and output of other operation nodes. The computational graph may clearly characterize the execution logic of at least part of the machine learning modeling process. Users can simply package own original codes into data nodes and operation nodes through a base class or a method provided by the framework, the data nodes and the operation nodes are connected into a calculation graph, and the connected calculation graph can automatically generate a relation graph. For example, when defining a computational graph, a user may use codes to simply package the logic of each operation node and a function (or class) provided by the framework, and the obtained multiple operation nodes may be connected according to the correspondence between the upstream and downstream and the input and output to obtain a computational graph, and the content of the computational graph is stored in the computational graph module. In addition, the process of generating the computational graph can be realized by means of a graphical interface, namely, a user can define nodes (operation nodes and data nodes) in the computational graph in a dragging mode on the graphical interface, the nodes can be existing logic provided by a framework, and the user can be supported to write a piece of custom code inside the nodes.
The parameter space module is configured to obtain a parameter space (corresponding to the above-mentioned value space of the operation parameter) configured by the user for the operation node in the computation graph. When a user defines an operation node in the computation graph, a sub-parameter space can be added to the operation node, the value of a certain operation parameter in the current operation node is changed from a determined item to an undetermined item, and one of the sub-parameter spaces can be selected from the value of the operation parameter during running. The sub-parameter spaces on all the operation nodes in the computation graph are combined, that is, the sub-parameter spaces correspond to the current computation graph, and generally, the product of all the sub-parameter space options forms the total parameter space. Taking the example of defining the computation graph by using the code by the user, the user may use a special identifier (such as a Choice class provided by the framework) for characterizing the sub-parameter space, and fill the special identifier in the position of the operation parameter related to the operation node, and the special identifier may be automatically identified when the computation graph is generated and added to the parameter space module for management. In addition, defining the sub-parameter space can also be done in a graphical interface, for example, a user can also define a parameter as the sub-parameter space using special options. In addition, for the operating parameters which usually have a fixed value space, the invention can also provide a default sub-parameter space to reduce the threshold used by the user, that is, for some operating parameters, the value space can be specified by the system without user definition.
Fig. 2 illustrates a computational graph for characterizing a machine-learned feature selection process. Where circles represent operational nodes and boxes represent data nodes.
The operation node read _1 in fig. 2 is an operation for reading data, and is used to read data from the hard disk into the memory; the data node variable _1 is used for representing original data in the memory; the operation node make _ dataset _1 is used for selecting k columns from the original data, and dividing the k columns into a training set and a test set according to rows, wherein the training set and the test set comprise a sub-parameter space (corresponding to the value space mentioned above), namely k columns of data of variable _1 are randomly selected; the data node variable _2 is used for characterizing a training data set only containing k columns; the data node variable _3 is used for characterizing a test data set only containing k columns; the operation node lgb _ train _1 is used to obtain a model on the training data set using the LightGBM, and calculate auc (accuracy) of the current model according to the test set, lgb _ train _1 contains two sub-parameter spaces inside, one is that the learning rate of the LightGBM is uniformly sampled between [0.1,0.2], and one is that the tree of the LightGBM trained tree is selected from [7,13,19,23 ]. Among them, the LightGBM is an abbreviation of the Light Gradient Boosting Machine, and is a framework for implementing the GBDT algorithm.
And the search strategy module is used for determining a plurality of groups of value combinations according to the calculation graph and the parameter space. When defining the computation graph, a user may define a target node, that is, an optimization target, where the optimization target is preferably a data node whose content is a scalar, and this node is a result of the operation logic represented by the current computation graph, and a target of the parameter search is that the result is optimal (maximum or minimum, generally, any evaluation method may be converted into a scalar). The search strategy module can calculate the next group or groups of parameter value combinations to be searched according to the historical parameter value combinations and the execution result combinations of the target nodes, and certain pruning optimization can be performed according to the existing algorithm in the process, so that the size of a search space is reduced, and the optimal solution is obtained in an accelerated manner.
After the search strategy module provides the next group or groups of parameter value combinations to be searched, the parameter value combinations and the original computation graph form a group of determined computation graph logics, then the execution framework can automatically schedule and execute the computation graph logics according to the machine resource conditions, calculate the execution result of the target node corresponding to the parameter value combinations, add the execution result to the historical information and use the history information to determine the parameter value combinations to be searched in the next round by the search strategy module.
After the execution engine and the search strategy module perform multiple iterations, when the search strategy module does not generate the next set of parameter value combination any more, the operation of the whole search process is finished, and then the parameter value combination corresponding to the best execution result in the historical data is the result obtained by the current search task.
Taking fig. 2 as an example, since the learning rate is a random value in a continuous space, the overall parameter space is infinite, and the search strategy may employ random search (random search), where each run randomly selects a value in each sub-parameter space, and after a total of M runs (random search parameters), the best result is selected from the random search. Wherein the numerical size of M is configurable by a user.
The code logic in the operation node can directly use the code logic in the original machine learning application of the user, and only needs to change some original fixed parameters into a sub-parameter space (namely, a value range) at a required position through simple packaging.
The method can realize the AutoML application only by slightly changing or learning few things, greatly reduces the development cost, does not depend on the specific logic in the operation node, and can be used by the user at will to realize any logic, even if a non-machine learning task can be run.
In addition, in view of execution efficiency, the search strategy module supports generation of multiple sets of parameter value combinations, generates multiple sets of calculation graphs and executes the calculation graphs simultaneously, so that the utilization rate of calculation resources is improved.
The execution engine is responsible for generating one or more computation graphs determined by the execution logic according to the value combination determined by the search strategy module, and operating the computation graphs to obtain an execution result (such as the value of the target node). In the execution process, the execution engine may be configured to determine which operation step or operation steps should be executed by each processing node in the distributed system, and the resource utilization efficiency in the distributed environment needs to be considered, and the resource utilization of an operation step may be simply predicted through historical execution information. For example, after multiple executions of an operation step, it is found that only one cpu is needed to meet its requirements, and then it can be calculated simultaneously with other intensive operation steps. Or the resource utilization rate of one calculation graph is not high all the time, the number of the calculation graphs returned by the search strategy module can be increased to increase the parallelism.
Search strategyThere may be intersections between different value combinations returned by the modules, for example, in the calculation diagram shown in fig. 2, only the parameter in the calculation node lgb _ train _1 is changed
Meanwhile, the output of the previous computing node can be reused without computation, and the execution engine can identify the node which can be skipped according to the parameter set and the current executed state and directly use the output of the node. The simultaneous execution engine may also release some resources in advance that are not used by the downstream nodes in the computation graph any more, according to the total resource situation (memory, cpu consumption, etc.) configured by the user.
In one embodiment of the invention, a machine learning model corresponding to at least a portion of the machine learning modeling process may be used in any of the following scenarios: online content (such as news, advertisements, music, etc.) recommendations; credit card fraud detection; detecting abnormal behaviors; intelligent marketing; an intelligent investment advisor; and analyzing the network traffic.
Still further, scenarios to which the machine learning model in embodiments of the present invention may be applied include, but are not limited to, the following scenarios:
an image processing scene comprising: optical character recognition OCR, face recognition, object recognition and picture classification; more specifically, for example, OCR may be applied to bill (e.g., invoice) recognition, handwritten character recognition, etc., face recognition may be applied to the fields of security, etc., object recognition may be applied to traffic sign recognition in an automatic driving scene, and picture classification may be applied to "buy by taking a picture", "find the same money", etc. of an e-commerce platform.
A voice recognition scene including products that can perform human-computer interaction through voice, such as a voice assistant of a mobile phone (e.g., Siri of an apple mobile phone), a smart sound box, and the like;
a natural language processing scenario, comprising: review text (e.g., contracts, legal documents, customer service records, etc.), spam content identification (e.g., spam short message identification), and text classification (sentiment, intent, subject matter, etc.);
an automatic control scenario, comprising: predicting mine group adjusting operation, predicting wind generating set adjusting operation and predicting air conditioning system adjusting operation; specifically, a group of adjustment operations with high predictable mining rate for a mine group, a group of adjustment operations with high predictable power generation efficiency for a wind generating set, and a group of adjustment operations with energy consumption saving while meeting requirements for an air conditioning system can be predicted;
an intelligent question-answering scenario comprising: a chat robot and an intelligent customer service;
a business decision scenario comprising: scene in finance science and technology field, medical field and municipal field, wherein:
the fields of financial science and technology include: marketing (e.g., coupon usage prediction, advertisement click behavior prediction, user portrait mining, etc.) and customer acquisition, anti-fraud, anti-money laundering, underwriting and credit scoring, commodity price prediction;
the medical field includes: disease screening and prevention, personalized health management and assisted diagnosis;
the municipal field includes: social administration and supervision law enforcement, resource environment and facility management, industrial development and economic analysis, public service and civil guarantee, and smart cities (allocation and management of various urban resources such as buses, online taxi appointment, shared bicycles, and the like);
recommending a business scenario, comprising: recommendations for news, advertisements, music, consultations, video, and financial products (e.g., financing, insurance, etc.);
searching for scenes, comprising: web page search, image search, text search, video search, and the like;
an abnormal behavior detection scenario comprising: the method comprises the steps of detecting abnormal power consumption behaviors of national grid customers, detecting network malicious flow, detecting abnormal behaviors in operation logs and the like.
The method for realizing automatic machine learning can also be realized as a device for realizing automatic machine learning. Fig. 3 illustrates a block diagram of an apparatus for implementing automatic machine learning according to an exemplary embodiment of the present invention. Wherein the functional elements of an apparatus for performing automatic machine learning are implemented in hardware, software, or a combination of hardware and software implementing the principles of the present invention. It will be appreciated by those skilled in the art that the functional units described in fig. 3 may be combined or divided into sub-units to implement the principles of the invention described above. Thus, the description herein may support any possible combination, or division, or further definition of the functional units described herein.
Functional units that can be provided by the apparatus for implementing automatic machine learning and operations that can be executed by each functional unit are briefly described below, and details related thereto may be referred to the above description, and are not repeated here.
Referring to fig. 3, the apparatus 300 for implementing automatic machine learning includes a first obtaining module 310, a second obtaining module 320, an executing module 330, and an outputting module 340.
The first obtaining module 310 is configured to obtain process configuration information set by a user for characterizing at least a portion of a machine learning modeling process, the process configuration information including one or more operational steps. The second obtaining module 320 is configured to obtain parameter configuration information set by a user for the at least part of the machine learning modeling process, where the parameter configuration information includes a value space of at least part of the operating parameters involved in the operating step. The specific implementation process of the first obtaining module 310 obtaining the process configuration information and the second obtaining module 320 obtaining the parameter configuration information may be as described above with reference to step S110 and step S120 in fig. 1.
As an example, the apparatus 300 implementing automatic machine learning may further include a providing module and a receiving module. The providing module is used for providing a data uploading interface for a user, and the receiving module is used for receiving a file uploaded by the user through the data uploading interface, wherein the file can be written by the user for at least part of the machine learning modeling process based on a language of a specific rule. The first acquisition module 310 may parse the file to determine process configuration information. Optionally, the file may further define a value space of at least part of the operation parameters involved in the operation step, and the second obtaining module 320 may parse the file to determine the parameter configuration information.
As an example, the apparatus 300 for implementing automatic machine learning may further include a first presentation module. The first presentation module is configured to present an interactive interface for setting the at least part of the machine learning modeling process to a user, where the first obtaining module 310 may obtain process configuration information set by the user through the interactive interface. The apparatus 300 for implementing automatic machine learning may further include a second presentation module. The second display module is configured to display, to a user, an interactive interface for setting a value space of the operation parameter related to the operation step, where the second obtaining module 320 may obtain parameter configuration information set by the user through the interactive interface.
The execution module 330 is configured to execute the at least part of the machine learning modeling process based on the process configuration information and different combinations of values of the at least part of the operating parameters determined based on the parameter configuration information, so as to obtain a plurality of execution results.
As an example, the process configuration information further defines at least one target node, the target node being a data node, and the apparatus 300 for implementing automatic machine learning may further include a determination module. In each iteration: the determining module may determine one or more sets of value combinations of the at least part of the operating parameters according to the parameter configuration information; the executing module 330 may execute the at least part of the machine learning modeling process according to the process configuration information and the value combination determined by the determining module, so as to obtain an execution result of the target node. Optionally, the apparatus 300 for implementing automatic machine learning may further comprise an add-on module. In each iteration: the adding module adds the value combination and the execution result determined by the current round to the historical information; and the determining module determines one or more groups of value combinations of at least part of the operation parameters in the current round by using a preset search strategy according to the parameter configuration information and the historical information. Wherein the predetermined search strategy may include, but is not limited to, at least one of: random search strategy, enumeration search strategy, Bayesian optimization search strategy and grid search strategy.
The apparatus 300 for implementing automatic machine learning may further include a saving module for saving an execution result of the operation step in which the value of the operation parameter is not changed. The executing module 330 may omit the execution of the operation step in which the value of the operation parameter is not changed in the at least part of the machine learning modeling process, and call the execution result of the operation step, which is pre-stored by the storing module, in the at least part of the machine learning modeling process.
The apparatus 300 for implementing automatic machine learning may further include a prediction module and an adjustment module. The estimation module is used for estimating the resource use condition of the computing resources distributed for the operation steps according to the execution result of at least part of the machine learning modeling process; and the adjusting module is used for adjusting the computing resources distributed for the operation steps according to the estimation result.
The output module 340 is configured to output a preferred value of the at least part of the operation parameters according to the execution results.
The apparatus 300 for implementing automatic machine learning may further include a third acquisition module and a filtering module. The third acquisition module is used for acquiring screening condition configuration information which is set by a user and used for screening at least part of the values of the operating parameters. The screening module is used for screening different value combinations of at least part of the operation parameters determined based on the parameter configuration information according to the screening condition configuration information. The execution module may execute at least a portion of the machine learning modeling process based on the process configuration information and the filtered value combinations. The screening condition configuration information may include one or more screening conditions. After the value combinations of at least part of the operating parameters are determined based on the parameter configuration information, the value combinations can be screened by the screening module according to the screening condition configuration information so as to eliminate the value combinations which do not meet the preset standard. The removed value combinations can be regarded as being unable to be solved and discarded, namely the removed value combinations do not participate in the execution of at least part of the machine learning modeling process any more.
It should be understood that, according to the embodiment of the present invention, the specific implementation manner of the apparatus 300 for implementing automatic machine learning can be implemented by referring to the above description related to fig. 1 for the method for implementing automatic machine learning, and is not described herein again.
A method and apparatus for implementing automatic machine learning according to exemplary embodiments of the present invention are described above with reference to fig. 1 to 3. It is to be understood that the above-described method may be implemented by a program recorded on a computer-readable medium, for example, according to an exemplary embodiment of the present invention, there may be provided a computer-readable storage medium storing instructions, wherein a computer program for executing the method of implementing automatic machine learning of the present invention (for example, shown in fig. 1) is recorded on the computer-readable medium.
The computer program in the computer-readable medium may be executed in an environment deployed in a computer device such as a client, a host, a proxy device, a server, etc., and it should be noted that the computer program may be used to perform additional steps or perform more specific processes when performing the steps in addition to or instead of the steps shown in fig. 1, and the contents of the additional steps and the further processes are described with reference to fig. 1, and will not be described again to avoid repetition.
It should be noted that the apparatus for implementing automatic machine learning according to the exemplary embodiment of the present invention may implement corresponding functions completely depending on the execution of the computer program, that is, each apparatus corresponds to each step in the functional architecture of the computer program, so that the whole apparatus is called by a special software package (e.g., lib library) to implement the corresponding functions.
Alternatively, the various means shown in fig. 3 may be implemented by hardware, software, firmware, middleware, microcode, or any combination thereof. When implemented in software, firmware, middleware or microcode, the program code or code segments to perform the corresponding operations may be stored in a computer-readable medium such as a storage medium, so that a processor may perform the corresponding operations by reading and executing the corresponding program code or code segments.
For example, exemplary embodiments of the present invention may also be implemented as a computing device comprising a storage component having stored therein a set of computer-executable instructions that, when executed by the processor, perform a method of implementing automatic machine learning.
In particular, the computing devices may be deployed in servers or clients, as well as on node devices in a distributed network environment. Further, the computing device may be a PC computer, tablet device, personal digital assistant, smart phone, web application, or other device capable of executing the set of instructions described above.
The computing device need not be a single computing device, but can be any device or collection of circuits capable of executing the instructions (or sets of instructions) described above, individually or in combination. The computing device may also be part of an integrated control system or system manager, or may be configured as a portable electronic device that interfaces with local or remote (e.g., via wireless transmission).
In the computing device, the processor may include a Central Processing Unit (CPU), a Graphics Processing Unit (GPU), a programmable logic device, a special purpose processor system, a microcontroller, or a microprocessor. By way of example, and not limitation, processors may also include analog processors, digital processors, microprocessors, multi-core processors, processor arrays, network processors, and the like.
Certain operations described in the method of implementing automatic machine learning according to exemplary embodiments of the present invention may be implemented by software, certain operations may be implemented by hardware, and further, the operations may be implemented by a combination of hardware and software.
The processor may execute instructions or code stored in one of the memory components, which may also store data. Instructions and data may also be transmitted and received over a network via a network interface device, which may employ any known transmission protocol.
The memory component may be integral to the processor, e.g., having RAM or flash memory disposed within an integrated circuit microprocessor or the like. Further, the storage component may comprise a stand-alone device, such as an external disk drive, storage array, or any other storage device usable by a database system. The storage component and the processor may be operatively coupled or may communicate with each other, such as through an I/O port, a network connection, etc., so that the processor can read files stored in the storage component.
Further, the computing device may also include a video display (such as a liquid crystal display) and a user interaction interface (such as a keyboard, mouse, touch input device, etc.). All components of the computing device may be connected to each other via a bus and/or a network.
Operations involved in methods of implementing automatic machine learning according to exemplary embodiments of the present invention may be described as various interconnected or coupled functional blocks or functional diagrams. However, these functional blocks or functional diagrams may be equally integrated into a single logic device or operated on by non-exact boundaries.
For example, as described above, an apparatus for implementing automatic machine learning according to an exemplary embodiment of the present invention may include a storage component and a processor, wherein the storage component stores therein a set of computer-executable instructions that, when executed by the processor, perform the above-mentioned method of implementing automatic machine learning.
While exemplary embodiments of the invention have been described above, it should be understood that the above description is illustrative only and not exhaustive, and that the invention is not limited to the exemplary embodiments disclosed. Many modifications and variations will be apparent to those of ordinary skill in the art without departing from the scope and spirit of the invention. Therefore, the protection scope of the present invention should be subject to the scope of the claims.
Claims (10)
1. A method of implementing automatic machine learning, comprising:
acquiring process configuration information set by a user and used for characterizing at least part of a machine learning modeling process, wherein the process configuration information comprises one or more operation steps;
acquiring parameter configuration information set by a user aiming at the at least part of machine learning modeling process, wherein the parameter configuration information comprises a value space of at least part of operation parameters related to the operation step;
executing the at least part of the machine learning modeling process based on the process configuration information and different value combinations of the at least part of the operating parameters determined based on the parameter configuration information to obtain a plurality of execution results; and
and outputting the optimal values of at least part of the operation parameters according to the execution results.
2. The method of claim 1, wherein obtaining user-set process configuration information characterizing at least a portion of a machine learning modeling process comprises:
providing a data uploading interface for a user;
receiving a file uploaded by a user through the data uploading interface, wherein the file is written by the user for at least part of the machine learning modeling process based on a language of a specific rule;
and analyzing the file to determine the process configuration information.
3. The method of claim 2, wherein the file further defines a value space of at least some of the operating parameters involved in the operating step, and the step of obtaining parameter configuration information set by a user for the at least some of the machine learning modeling processes includes:
and analyzing the file to determine the parameter configuration information.
4. The method of claim 1, wherein obtaining user-set process configuration information characterizing at least a portion of a machine learning modeling process comprises:
presenting an interactive interface to a user for setting up the at least part of the machine learning modeling process;
and acquiring process configuration information set by a user through the interactive interface.
5. The method of claim 1, wherein obtaining parameter configuration information set by a user for the at least partial machine learning modeling process comprises:
displaying an interactive interface for setting a value space of the operation parameters related to the operation steps to a user;
and acquiring parameter configuration information set by the user through the interactive interface.
6. The method of claim 1, wherein the process configuration information further defines at least one target node, the target node being a data node, wherein the step of performing the at least partial machine learning modeling process based on the process configuration information and different combinations of values of the at least partial operating parameter determined based on the parameter configuration information comprises iteratively performing the following:
determining one or more groups of value combinations of the at least part of the operating parameters according to the parameter configuration information;
and executing at least part of the machine learning modeling process according to the process configuration information and the determined value combination to obtain an execution result of the target node.
7. The method of claim 6, wherein during each iteration, the method further comprises: adding the value combination and execution result determined by the current round to the historical information,
the step of determining one or more sets of value combinations of the at least part of the operating parameters according to the parameter configuration information comprises: and determining one or more groups of value combinations of the at least part of the operation parameters in the current round by using a preset search strategy according to the parameter configuration information and the historical information.
8. An apparatus to implement automatic machine learning, comprising:
the system comprises a first acquisition module, a second acquisition module and a third acquisition module, wherein the first acquisition module is used for acquiring process configuration information which is set by a user and is used for characterizing at least part of a machine learning modeling process, and the process configuration information comprises one or more operation steps;
a second obtaining module, configured to obtain parameter configuration information set by a user for the at least part of machine learning modeling process, where the parameter configuration information includes a value space of at least part of operation parameters involved in the operation step;
an execution module, configured to execute the at least part of the machine learning modeling process based on the process configuration information and different value combinations of the at least part of the operating parameters determined based on the parameter configuration information, to obtain a plurality of execution results; and
and the output module is used for outputting the optimal values of at least part of the operating parameters according to the execution results.
9. A system comprising at least one computing device and at least one storage device storing instructions that, when executed by the at least one computing device, cause the at least one computing device to perform the method of any of claims 1 to 7.
10. A computer-readable storage medium storing instructions that, when executed by at least one computing device, cause the at least one computing device to perform the method of any of claims 1 to 7.
Priority Applications (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010219796.5A CN111340240A (en) | 2020-03-25 | 2020-03-25 | Method and device for realizing automatic machine learning |
| PCT/CN2021/081329 WO2021190379A1 (en) | 2020-03-25 | 2021-03-17 | Method and device for realizing automatic machine learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010219796.5A CN111340240A (en) | 2020-03-25 | 2020-03-25 | Method and device for realizing automatic machine learning |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN111340240A true CN111340240A (en) | 2020-06-26 |
Family
ID=71186282
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010219796.5A Pending CN111340240A (en) | 2020-03-25 | 2020-03-25 | Method and device for realizing automatic machine learning |
Country Status (2)
| Country | Link |
|---|---|
| CN (1) | CN111340240A (en) |
| WO (1) | WO2021190379A1 (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112783478A (en) * | 2021-02-19 | 2021-05-11 | 合肥海赛信息科技有限公司 | Software design method based on automatic machine learning |
| CN113033816A (en) * | 2021-03-08 | 2021-06-25 | 北京沃东天骏信息技术有限公司 | Processing method and device of machine learning model, storage medium and electronic equipment |
| WO2021190379A1 (en) * | 2020-03-25 | 2021-09-30 | 第四范式（北京）技术有限公司 | Method and device for realizing automatic machine learning |
| CN114385256A (en) * | 2020-10-22 | 2022-04-22 | 华为云计算技术有限公司 | Method and device for configuring system parameters |
Families Citing this family (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114528477A (en) * | 2022-01-10 | 2022-05-24 | 华南理工大学 | Scientific research application-oriented automatic machine learning implementation method, platform and device |
| CN115334005B (en) * | 2022-03-31 | 2024-03-22 | 北京邮电大学 | Encryption flow identification method based on pruning convolutional neural network and machine learning |
Citations (18)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106326005A (en) * | 2016-01-21 | 2017-01-11 | 华南师范大学 | Automatic parameter tuning method for iterative MapReduce operation |
| CN107247973A (en) * | 2017-06-29 | 2017-10-13 | 中国矿业大学 | A kind of preferred Parallel Particle Swarm Optimization optimization method of SVMs parameter based on spark |
| CN108446299A (en) * | 2018-01-25 | 2018-08-24 | 链家网（北京）科技有限公司 | The method and device of data-optimized calculating in a kind of task |
| CN108595265A (en) * | 2018-04-11 | 2018-09-28 | 武汉唯信兄弟科技有限公司 | Intelligent distribution method and system for computing resources |
| CN108632054A (en) * | 2017-03-17 | 2018-10-09 | 电子科技大学 | The prediction technique and device of information propagation amount |
| CN108710949A (en) * | 2018-04-26 | 2018-10-26 | 第四范式（北京）技术有限公司 | The method and system of template are modeled for creating machine learning |
| CN108845874A (en) * | 2018-06-25 | 2018-11-20 | 腾讯科技（深圳）有限公司 | The dynamic allocation method and server of resource |
| CN109242105A (en) * | 2018-08-17 | 2019-01-18 | 第四范式（北京）技术有限公司 | Tuning method, apparatus, equipment and the medium of hyper parameter in machine learning model |
| CN109464803A (en) * | 2018-11-05 | 2019-03-15 | 腾讯科技（深圳）有限公司 | Virtual objects controlled, model training method, device, storage medium and equipment |
| CN110166312A (en) * | 2018-02-16 | 2019-08-23 | 丛林网络公司 | Network equipment model is automatically created using random test bed |
| CN110321222A (en) * | 2019-07-01 | 2019-10-11 | 中国人民解放军国防科技大学 | Decision tree prediction-based data parallel operation resource allocation method |
| CN110472747A (en) * | 2019-08-16 | 2019-11-19 | 第四范式（北京）技术有限公司 | For executing the distributed system and its method of multimachine device learning tasks |
| CN110503208A (en) * | 2019-08-26 | 2019-11-26 | 第四范式（北京）技术有限公司 | Resource regulating method and resource scheduling device in multi-model exploration |
| CN110598868A (en) * | 2018-05-25 | 2019-12-20 | 腾讯科技（深圳）有限公司 | Machine learning model building method and device and related equipment |
| US20190394466A1 (en) * | 2018-06-25 | 2019-12-26 | Tfi Digital Media Limited | Method for initial quantization parameter optimization in video coding |
| AU2019101182A4 (en) * | 2019-10-02 | 2020-01-23 | Feng, Yawen MISS | Credit Risk Assessment of Lending Borrowers Based on Hybrid Supervised and Unsupervised Learning |
| CN110766164A (en) * | 2018-07-10 | 2020-02-07 | 第四范式（北京）技术有限公司 | Method and system for performing a machine learning process |
| CN110895718A (en) * | 2018-09-07 | 2020-03-20 | 第四范式（北京）技术有限公司 | Method and system for training machine learning model |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111340240A (en) * | 2020-03-25 | 2020-06-26 | 第四范式（北京）技术有限公司 | Method and device for realizing automatic machine learning |
-
2020
- 2020-03-25 CN CN202010219796.5A patent/CN111340240A/en active Pending
-
2021
- 2021-03-17 WO PCT/CN2021/081329 patent/WO2021190379A1/en active Application Filing
Patent Citations (18)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106326005A (en) * | 2016-01-21 | 2017-01-11 | 华南师范大学 | Automatic parameter tuning method for iterative MapReduce operation |
| CN108632054A (en) * | 2017-03-17 | 2018-10-09 | 电子科技大学 | The prediction technique and device of information propagation amount |
| CN107247973A (en) * | 2017-06-29 | 2017-10-13 | 中国矿业大学 | A kind of preferred Parallel Particle Swarm Optimization optimization method of SVMs parameter based on spark |
| CN108446299A (en) * | 2018-01-25 | 2018-08-24 | 链家网（北京）科技有限公司 | The method and device of data-optimized calculating in a kind of task |
| CN110166312A (en) * | 2018-02-16 | 2019-08-23 | 丛林网络公司 | Network equipment model is automatically created using random test bed |
| CN108595265A (en) * | 2018-04-11 | 2018-09-28 | 武汉唯信兄弟科技有限公司 | Intelligent distribution method and system for computing resources |
| CN108710949A (en) * | 2018-04-26 | 2018-10-26 | 第四范式（北京）技术有限公司 | The method and system of template are modeled for creating machine learning |
| CN110598868A (en) * | 2018-05-25 | 2019-12-20 | 腾讯科技（深圳）有限公司 | Machine learning model building method and device and related equipment |
| CN108845874A (en) * | 2018-06-25 | 2018-11-20 | 腾讯科技（深圳）有限公司 | The dynamic allocation method and server of resource |
| US20190394466A1 (en) * | 2018-06-25 | 2019-12-26 | Tfi Digital Media Limited | Method for initial quantization parameter optimization in video coding |
| CN110766164A (en) * | 2018-07-10 | 2020-02-07 | 第四范式（北京）技术有限公司 | Method and system for performing a machine learning process |
| CN109242105A (en) * | 2018-08-17 | 2019-01-18 | 第四范式（北京）技术有限公司 | Tuning method, apparatus, equipment and the medium of hyper parameter in machine learning model |
| CN110895718A (en) * | 2018-09-07 | 2020-03-20 | 第四范式（北京）技术有限公司 | Method and system for training machine learning model |
| CN109464803A (en) * | 2018-11-05 | 2019-03-15 | 腾讯科技（深圳）有限公司 | Virtual objects controlled, model training method, device, storage medium and equipment |
| CN110321222A (en) * | 2019-07-01 | 2019-10-11 | 中国人民解放军国防科技大学 | Decision tree prediction-based data parallel operation resource allocation method |
| CN110472747A (en) * | 2019-08-16 | 2019-11-19 | 第四范式（北京）技术有限公司 | For executing the distributed system and its method of multimachine device learning tasks |
| CN110503208A (en) * | 2019-08-26 | 2019-11-26 | 第四范式（北京）技术有限公司 | Resource regulating method and resource scheduling device in multi-model exploration |
| AU2019101182A4 (en) * | 2019-10-02 | 2020-01-23 | Feng, Yawen MISS | Credit Risk Assessment of Lending Borrowers Based on Hybrid Supervised and Unsupervised Learning |
Non-Patent Citations (2)
| Title |
|---|
| BISHWAJIT ROY ET AL: "A Metaheuristic-based Emotional ANN (EmNN) Approach for Rainfall-runoff Modeling", 《2019 INTERNATIONAL CONFERENCE ON COMMUNICATION AND ELECTRONICS SYSTEMS (ICCES)》 * |
| 车江涛: "基于Spark的智慧城市房价评估系统的研究与实现", 《中国优秀硕士学位论文全文数据库 信息科技辑》 * |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2021190379A1 (en) * | 2020-03-25 | 2021-09-30 | 第四范式（北京）技术有限公司 | Method and device for realizing automatic machine learning |
| CN114385256A (en) * | 2020-10-22 | 2022-04-22 | 华为云计算技术有限公司 | Method and device for configuring system parameters |
| CN112783478A (en) * | 2021-02-19 | 2021-05-11 | 合肥海赛信息科技有限公司 | Software design method based on automatic machine learning |
| CN113033816A (en) * | 2021-03-08 | 2021-06-25 | 北京沃东天骏信息技术有限公司 | Processing method and device of machine learning model, storage medium and electronic equipment |
Also Published As
| Publication number | Publication date |
|---|---|
| WO2021190379A1 (en) | 2021-09-30 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| WO2020253775A1 (en) | Method and system for realizing machine learning modeling process | |
| WO2020249125A1 (en) | Method and system for automatically training machine learning model | |
| WO2021190379A1 (en) | Method and device for realizing automatic machine learning | |
| CN111523677B (en) | Method and device for realizing interpretation of prediction result of machine learning model | |
| US20210264272A1 (en) | Training method and system of neural network model and prediction method and system | |
| CN110705719A (en) | Method and apparatus for performing automatic machine learning | |
| CN110717597A (en) | Method and device for acquiring time sequence characteristics by using machine learning model | |
| US20230043882A1 (en) | Method for assisting launch of machine learning model | |
| CN111651524B (en) | Auxiliary implementation method and device for on-line prediction by using machine learning model | |
| CN111966886A (en) | Object recommendation method, object recommendation device, electronic equipment and storage medium | |
| CN112328869A (en) | User loan willingness prediction method and device and computer system | |
| CN111768096A (en) | Rating method and device based on algorithm model, electronic equipment and storage medium | |
| US20210326761A1 (en) | Method and System for Uniform Execution of Feature Extraction | |
| CN114862140A (en) | Behavior analysis-based potential evaluation method, device, equipment and storage medium | |
| CN108229572B (en) | Parameter optimization method and computing equipment | |
| CN110544166A (en) | Sample generation method, device and storage medium | |
| CN115641186A (en) | Intelligent analysis method, device and equipment for preference of live broadcast product and storage medium | |
| CN111859985B (en) | AI customer service model test method and device, electronic equipment and storage medium | |
| CN114003567A (en) | Data acquisition method and related device | |
| CN111178535B (en) | Method and apparatus for implementing automatic machine learning | |
| US20220398433A1 (en) | Efficient Cross-Platform Serving of Deep Neural Networks for Low Latency Applications | |
| CN117541884A (en) | Sample data processing method, device, storage medium and system | |
| CN117541885A (en) | Sample data processing method, device, storage medium and system | |
| CN117539857A (en) | Table splicing method, apparatus, storage medium and system | |
| CN116611923A (en) | Knowledge graph-based risk data acquisition method, system, equipment and storage medium |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |