CN116959629A - Multi-index optimization method and system for chemical experiment, storage medium and electronic equipment - Google Patents
Multi-index optimization method and system for chemical experiment, storage medium and electronic equipment Download PDFInfo
- Publication number
- CN116959629A CN116959629A CN202311218216.0A CN202311218216A CN116959629A CN 116959629 A CN116959629 A CN 116959629A CN 202311218216 A CN202311218216 A CN 202311218216A CN 116959629 A CN116959629 A CN 116959629A
- Authority
- CN
- China
- Prior art keywords
- experimental
- value
- index
- experiment
- values
- 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.)
- Granted
Links
- 238000002474 experimental method Methods 0.000 title claims abstract description 140
- 238000005457 optimization Methods 0.000 title claims abstract description 82
- 238000000034 method Methods 0.000 title claims abstract description 69
- 239000000126 substance Substances 0.000 title claims abstract description 56
- 238000003860 storage Methods 0.000 title abstract description 7
- 238000004422 calculation algorithm Methods 0.000 claims abstract description 37
- 230000008569 process Effects 0.000 claims abstract description 20
- 238000011423 initialization method Methods 0.000 claims abstract description 8
- 238000013507 mapping Methods 0.000 claims description 43
- 238000012545 processing Methods 0.000 claims description 20
- 238000010276 construction Methods 0.000 claims description 11
- 238000009826 distribution Methods 0.000 claims description 11
- 230000006872 improvement Effects 0.000 claims description 11
- 238000004364 calculation method Methods 0.000 claims description 8
- 238000010606 normalization Methods 0.000 claims description 6
- 238000012163 sequencing technique Methods 0.000 claims description 5
- 238000004519 manufacturing process Methods 0.000 abstract description 5
- 238000012795 verification Methods 0.000 abstract description 3
- 230000006870 function Effects 0.000 description 13
- 239000003054 catalyst Substances 0.000 description 9
- 238000010586 diagram Methods 0.000 description 7
- 238000006243 chemical reaction Methods 0.000 description 3
- 239000003795 chemical substances by application Substances 0.000 description 3
- 238000004590 computer program Methods 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000004458 analytical method Methods 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 238000013500 data storage Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 230000010354 integration Effects 0.000 description 1
- 230000003993 interaction Effects 0.000 description 1
- 239000000463 material Substances 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 239000013307 optical fiber Substances 0.000 description 1
- 238000007781 pre-processing Methods 0.000 description 1
- 239000004065 semiconductor Substances 0.000 description 1
- 239000007787 solid Substances 0.000 description 1
Classifications
-
- G—PHYSICS
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16C—COMPUTATIONAL CHEMISTRY; CHEMOINFORMATICS; COMPUTATIONAL MATERIALS SCIENCE
- G16C20/00—Chemoinformatics, i.e. ICT specially adapted for the handling of physicochemical or structural data of chemical particles, elements, compounds or mixtures
- G16C20/70—Machine learning, data mining or chemometrics
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N7/00—Computing arrangements based on specific mathematical models
- G06N7/01—Probabilistic graphical models, e.g. probabilistic networks
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
- Y02P90/00—Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
- Y02P90/30—Computing systems specially adapted for manufacturing
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Software Systems (AREA)
- Artificial Intelligence (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Computing Systems (AREA)
- General Health & Medical Sciences (AREA)
- Probability & Statistics with Applications (AREA)
- Crystallography & Structural Chemistry (AREA)
- Life Sciences & Earth Sciences (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Medical Informatics (AREA)
- Health & Medical Sciences (AREA)
- Databases & Information Systems (AREA)
- Chemical & Material Sciences (AREA)
- Algebra (AREA)
- Computational Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Mathematical Analysis (AREA)
- Mathematical Optimization (AREA)
- Pure & Applied Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Mathematical Physics (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
A multi-index optimization method, a system, a storage medium and an electronic device for chemical experiment adopt a differentiated data initialization method according to the existence of historical data of experimental information of experimental factors in a designated aspect; taking the initialized historical data consisting of the numerical tensors as the input of a Bayesian optimization algorithm, and recommending a group of non-repeated experimental factor parameter combination schemes through Gaussian process regression and preference comparison selection; and respectively carrying out experiments on the recommended experimental factor parameter combination schemes of the previous round, merging the recommended experimental factor parameter combination schemes with the obtained corresponding index values, adding the merged experimental factor parameter combination schemes to a historical data set, taking all historical data as the input of a Bayesian optimization algorithm of the next round, recommending a set of experimental factor parameter combination schemes for experiments again, and carrying out a plurality of rounds of loop iteration on the Bayesian optimization algorithm by taking the condition that all the index values are simultaneously in a preset range as stop conditions. The invention omits tedious and costly chemical experiment verification, and can effectively reduce the production cost.
Description
Technical Field
The invention belongs to the technical field of chemical analysis, and particularly relates to a multi-index optimization method and system for chemical experiments.
Background
Bayesian optimization (Bayesian Optimization) is a method of super-parametric searching, which uses gaussian process regression (gaussian process regression) to model probability distribution of existing histories, and iteratively performs super-parametric selection.
At present, in the chemical field, the actual chemical experiment cost is extremely high, a large amount of manpower and material resources are consumed, the search space is huge, and the total parameter combination is difficult to enumerate; in practical situations, there may be multiple indexes (multi-objects) at the same time, such as yield, cost, and the like, where there is an inherent conflict between indexes, and the cost of optimizing one index is to degrade other indexes, so that a unique optimal solution is difficult to occur, and the complexity of the problem is significantly improved due to the increase of the number of indexes to be optimized, which is difficult to be qualified by a conventional bayesian optimization algorithm.
In addition, since chemical experiments involve complex chemical reaction processes that do not follow explicit functional relationships and thus cannot be resolved analytically, multiple index values for these schemes need to be obtained by performing actual chemical experiments. Therefore, how to solve the multi-index optimization problem in the chemical industry by using bayesian optimization to recommend a preferred scheme to make the staff perform the complex experiment as little as possible is a technical problem to be solved at present.
Disclosure of Invention
Therefore, the invention provides a multi-index optimization method, a system, a storage medium and electronic equipment for chemical experiments, which solve the problem of multi-index optimization in the chemical industry so as to recommend a preferred scheme and enable workers to carry out complex experiments as little as possible.
In order to achieve the above object, the present invention provides the following technical solutions: a method for optimizing multiple indexes of a chemical experiment, comprising:
s1: experiment information configuration and data initialization:
s11: determining an index to be optimized and experimental information of experimental factors in a specified aspect;
s12: according to the existence of historical data of experimental information of experimental factors in the appointed aspect, adopting a differentiated data initialization method;
s2: reading in the initialized experimental information data, and carrying out iterative recommendation:
s21: taking the initialized historical data consisting of the numerical tensors as the input of a Bayesian optimization algorithm, and recommending a group of non-repeated experimental factor parameter combination schemes through Gaussian process regression and preference comparison selection;
s22: and respectively carrying out experiments on the recommended experimental factor parameter combination schemes of the previous round, merging the recommended experimental factor parameter combination schemes with the obtained corresponding index values, adding the merged experimental factor parameter combination schemes to a historical data set, taking all historical data as the input of a Bayesian optimization algorithm of the next round, recommending a set of experimental factor parameter combination schemes for experiments again, and carrying out a plurality of rounds of loop iteration on the Bayesian optimization algorithm by taking the condition that all the index values are simultaneously in a preset range as stop conditions.
As a preferred embodiment of the chemical experiment multi-index optimization method, step S11 includes:
s111: the method comprises the steps of performing permutation and combination on a current chemical experiment by using the names and the value types of experimental factors in the appointed aspect and the parameter values of the experimental factors in the appointed aspect to construct a full search space formed by mixing numerical type values and character type values;
s112: designating the name of the index to be optimized, designating the expected optimization direction of the index to be optimized, and designating the weight corresponding to the index to be optimized.
As a preferred embodiment of the chemical experiment multi-index optimization method, step S111 includes:
s1111: sequencing all values of the experimental factors to form an ordered list from small to large, and forming a nested list consisting of a corresponding number of ordered lists if more than one numerical experimental factor exists;
s1112: and mapping and encoding each different value to form a mapping dictionary taking the encoded number as a key and taking the character string as a value, and if more than one character string type experimental factors exist, forming a nested list consisting of a corresponding number of mapping dictionaries.
As a preferred embodiment of the chemical experiment multi-index optimization method, step S12 includes:
s121: if an experimental scheme and a corresponding experimental index value exist, the experimental scheme and the corresponding experimental index value are directly used as a historical data set;
s122: if no experiment record exists, randomly generating a preset number of non-repeated experiment schemes from the search space in the step S111 to serve as initial samples, then carrying out experiments on the experiment schemes to obtain index values actually corresponding to each experiment scheme, and splicing the experiment schemes and the corresponding index values to obtain a historical data set.
As a preferred embodiment of the chemical experiment multi-index optimization method, step S21 includes:
s211: converting an original value into a tensor by using Pytorch as input to the value type value history data; for the string type value, firstly mapping the string value into a number corresponding to the code according to the mapping dictionary constructed in the step S1112, and then converting the number corresponding to the code into a tensor by using Pytorch as input; and converting the index value into a tensor using Pytorch;
s212: combining the numerical model value and the tensor converted by the character string model value, carrying out normalization operation on parameters, constructing a probability proxy model, fitting priori knowledge through Gaussian process regression to obtain approximate distribution of real problems, optimizing an acquisition function EI, and taking out experimental parameter combinations which enable the index improvement degree of the next round to be the largest as candidate samples.
As a preferred embodiment of the chemical experiment multi-index optimization method, step S22 includes:
s221: the tensor composed of numerical values output by the Bayesian optimization algorithm is used as a column value, and the names of all experimental factors are used as column names to construct a data frame;
s222: outputting the nearest experimental factor value by utilizing the ordered list generated in the step S1111 aiming at the numerical value column;
s223: for the string type column, firstly, rounding operation is performed on each value by a round method in the Python scientific calculation package numpy, then, an integer obtained by the rounding operation is used as a key, the mapping dictionary constructed in the step S1112 is searched, and the string value corresponding to the key is taken out.
The invention also provides a chemical experiment multi-index optimization system, which comprises:
the experimental information configuration module is used for determining the to-be-optimized index and the experimental information of the experimental factors in the appointed aspect;
the data initialization module is used for adopting a differential data initialization method according to whether historical data exists in experimental information of experimental factors in the appointed aspect;
the experimental combination scheme recommending module is used for recommending a group of non-repeated experimental factor parameter combination schemes by taking the initialized historical data consisting of the numerical tensor as the input of a Bayesian optimization algorithm and through Gaussian process regression and preference comparison and selection;
And the experiment combination scheme loop iteration module is used for respectively carrying out experiments on the experiment factor parameter combination schemes recommended in the previous round, merging the experiment factor parameter combination schemes with the obtained corresponding index values, adding the merged index values to a historical data set, taking all the historical data as the input of the Bayesian optimization algorithm in the next round, recommending a group of experiment factor parameter combination schemes again for the experiments, and carrying out loop iteration on the Bayesian optimization algorithm for a plurality of rounds by taking the condition that all the index values are simultaneously in a preset range.
As a preferred scheme of the chemical experiment multi-index optimization system, the experiment information configuration module comprises:
the search space construction submodule is used for carrying out permutation and combination on the current chemical experiment by utilizing the names and the value types of the experimental factors in the appointed aspect and the parameter values of the experimental factors in the appointed aspect to construct a full amount of search space formed by mixing the numerical type values and the character type values;
the index parameter setting submodule is used for designating the name of the index to be optimized, designating the expected optimization direction of the index to be optimized and designating the weight corresponding to the index to be optimized;
the search space construction submodule includes:
The numerical value processing submodule is used for sequencing all values of experimental factors consisting of numerical values to form an ordered list from small to large, and if more than one numerical experimental factor exists, a nested list consisting of the ordered list with corresponding number is formed;
the character string type value processing sub-module is used for mapping and encoding each different value of experimental factors consisting of character string type values to form a mapping dictionary taking an encoded number as a key and taking a character string as a value, and if more than one character string type experimental factor exists, a nested list consisting of a corresponding number of mapping dictionaries is formed;
the data initialization module comprises:
the experimental index value calling sub-module is used for directly using the experimental index value as a historical data set if an experimental scheme and a corresponding experimental index value exist;
and the experiment index value generation sub-module is used for randomly generating a preset number of non-repeated experiment schemes in the search space of the search space construction sub-module as initial samples if no experiment record exists, carrying out experiments on the experiment schemes to obtain index values actually corresponding to each experiment scheme, and splicing the experiment schemes and the corresponding index values to obtain a historical data set.
As a preferred scheme of the chemical experiment multi-index optimization system, the experiment combination scheme recommendation module comprises:
the tensor generation submodule is used for converting an original value into a tensor by using Pytorch as input to the numerical value historical data; the mapping dictionary is also used for mapping the character string type value into numbers corresponding to codes according to the mapping dictionary constructed by the character string type value processing submodule, and then converting the numbers corresponding to the codes into tensor by using Pytorch as input; and converting the index value into a tensor using Pytorch;
the experimental parameter candidate sample generation sub-module is used for combining the tensor converted by the numerical model value and the character string model value, carrying out normalization operation on the parameters, constructing a probability agent model, fitting priori knowledge through Gaussian process regression to obtain approximate distribution of real problems, optimizing an acquisition function EI, and taking out the experimental parameter combination which enables the index improvement degree of the next round to be the largest as a candidate sample.
As a preferred scheme of the chemical experiment multi-index optimization system, the experiment combination scheme loop iteration module comprises:
the data frame construction submodule is used for constructing a data frame by taking a tensor consisting of numerical values output by a Bayesian optimization algorithm as a column value and taking the names of all experimental factors as column names;
The experimental factor value output sub-module is used for outputting the experimental factor value closest to the ordered list generated by the numerical value processing sub-module aiming at the numerical value column;
the rounding operation processing sub-module is used for carrying out rounding operation on each value by a round method in the Python scientific calculation packet numpy aiming at the character string type column, and searching a mapping dictionary constructed by the character string type value processing sub-module by taking an integer obtained by the rounding operation as a key to take out a character string value corresponding to the key.
The invention has the following advantages: determining an index to be optimized and experimental information of experimental factors in a specified aspect; according to the existence of historical data of experimental information of experimental factors in the appointed aspect, adopting a differentiated data initialization method; taking the initialized historical data consisting of the numerical tensors as the input of a Bayesian optimization algorithm, and recommending a group of non-repeated experimental factor parameter combination schemes through Gaussian process regression and preference comparison selection; and respectively carrying out experiments on the recommended experimental factor parameter combination schemes of the previous round, merging the recommended experimental factor parameter combination schemes with the obtained corresponding index values, adding the merged experimental factor parameter combination schemes to a historical data set, taking all historical data as the input of a Bayesian optimization algorithm of the next round, recommending a set of experimental factor parameter combination schemes for experiments again, and carrying out a plurality of rounds of loop iteration on the Bayesian optimization algorithm by taking the condition that all the index values are simultaneously in a preset range as stop conditions. According to the invention, as experimental data of multiple iterations are accumulated, the probability agent model can be more approximate to a real distribution curve, and at the moment, the optimized acquisition function can reasonably explore and utilize the residual area according to the curve, so that the target is continuously close to the set direction, and a scheme for combining experimental factor parameters, which enables multiple index values to reach better simultaneously, can be obtained after limited iterations, cumbersome and costly chemical experiment verification is omitted, the production cost can be effectively reduced, and the reasonable selection problem of experimental conditions can be effectively solved by related technicians.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below. It will be apparent to those skilled in the art from this disclosure that the drawings described below are merely exemplary and that other embodiments may be derived from the drawings provided without undue effort.
FIG. 1 is a schematic overall flow chart of a chemical experiment multi-index optimization method provided in an embodiment of the invention;
FIG. 2 is a schematic diagram of an implementation flow of step S1 in a multi-index optimization method for chemical experiments according to an embodiment of the present invention;
FIG. 3 is a schematic diagram of an implementation flow of step S2 in a multi-index optimization method for chemical experiments according to an embodiment of the present invention;
FIG. 4 is a schematic flow chart of the step S11 of the chemical experiment multi-index optimization method according to the embodiment of the invention;
FIG. 5 is a schematic diagram of an implementation flow of step S111 of a multi-index optimization method for chemical experiments according to an embodiment of the present invention;
FIG. 6 is a schematic diagram of an implementation flow of step S12 of a multi-index optimization method for chemical experiments according to an embodiment of the present invention;
FIG. 7 is a schematic diagram of an implementation flow of step S21 of a multi-index optimization method for chemical experiments according to an embodiment of the present invention;
FIG. 8 is a schematic diagram of an implementation flow of step S22 of a multi-index optimization method for chemical experiments according to an embodiment of the present invention;
fig. 9 is an algorithm schematic diagram of a chemical experiment multi-index optimization method provided in an embodiment of the invention.
Detailed Description
Other advantages and advantages of the present invention will become apparent to those skilled in the art from the following detailed description, which, by way of illustration, is to be read in connection with certain specific embodiments, but not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
Example 1:
referring to fig. 1, 2 and 3, embodiment 1 of the present invention provides a chemical experiment multi-index optimization method, which includes the following steps:
s1: experiment information configuration and data initialization:
s11: determining an index to be optimized and experimental information of experimental factors in a specified aspect;
S12: according to the existence of historical data of experimental information of experimental factors in the appointed aspect, adopting a differentiated data initialization method;
s2: reading in the initialized experimental information data, and carrying out iterative recommendation:
s21: taking the initialized historical data consisting of the numerical tensors as the input of a Bayesian optimization algorithm, and recommending a group of non-repeated experimental factor parameter combination schemes through Gaussian process regression and preference comparison selection;
s22: and respectively carrying out experiments on the recommended experimental factor parameter combination schemes of the previous round, merging the recommended experimental factor parameter combination schemes with the obtained corresponding index values, adding the merged experimental factor parameter combination schemes to a historical data set, taking all historical data as the input of a Bayesian optimization algorithm of the next round, recommending a set of experimental factor parameter combination schemes for experiments again, and carrying out a plurality of rounds of loop iteration on the Bayesian optimization algorithm by taking the condition that all the index values are simultaneously in a preset range as stop conditions.
Referring to fig. 4 and 5, in the present embodiment, step S11 includes:
s111: the method comprises the steps of performing permutation and combination on a current chemical experiment by using the names and the value types of experimental factors in the appointed aspect and the parameter values of the experimental factors in the appointed aspect to construct a full search space formed by mixing numerical type values and character type values;
S112: designating the name of the index to be optimized, designating the expected optimization direction of the index to be optimized, and designating the weight corresponding to the index to be optimized.
In step S112, the desired optimization direction of the to-be-optimized index is generally that it is desired to maximize the yield and minimize the cost, and the weight corresponding to the to-be-optimized index represents the importance of the index, and the larger the weight is, the higher the importance of the to-be-optimized index is.
Wherein, step S111 includes:
s1111: sequencing all values of the experimental factors to form an ordered list from small to large, and forming a nested list consisting of a corresponding number of ordered lists if more than one numerical experimental factor exists;
s1112: and mapping and encoding each different value to form a mapping dictionary taking the encoded number as a key and taking the character string as a value, and if more than one character string type experimental factors exist, forming a nested list consisting of a corresponding number of mapping dictionaries.
Referring to fig. 6, in the present embodiment, step S12 includes:
s121: if an experimental scheme and a corresponding experimental index value exist, the experimental scheme and the corresponding experimental index value are directly used as a historical data set;
S122: if no experiment record exists, randomly generating a preset number of non-repeated experiment schemes from the search space in the step S111 to serve as initial samples, then carrying out experiments on the experiment schemes to obtain index values actually corresponding to each experiment scheme, and splicing the experiment schemes and the corresponding index values to obtain a historical data set.
Specifically, in step S1 of this embodiment, experimental factors of specific chemical reactions and their respective values are determined, for example, three experimental factors of temperature, pressure and catalyst are selected, wherein the temperature and pressure values are numerical values, the temperature can be 10,20,30, the pressure can be 1,2,3, the catalyst values are character strings, sample 1, sample 2, sample 3 can be taken, and the size of the search space formed by the above steps is 3×3×3=27. Forming a nested ordered list of values, [ [10,20,30], [1,2,3] ], for temperature and pressure; for the catalyst, a key value mapping dictionary {0: 'sample 1',1: 'sample 2',2: 'sample 3'. The index to be optimized is determined, here taking the double index as an example, namely the yield and the cost, and the cost is minimized while the yield is expected to be maximized, and the two are given the same weight.
Wherein, under the condition of historical data, the original value can be directly converted into a tensor by using Pytorch as input for temperature and pressure; for the catalyst, mapping the catalyst into numbers corresponding to codes, converting the numbers into tensor tensors by using Pytorch, performing tensor conversion on index values, uniformly normalizing parameters, and performing standardization treatment on the index; under the condition of no historical data, a group of experimental parameter combinations need to be initialized randomly from the search space, after the experimental results are obtained, the corresponding multi-index values are obtained, and then the preprocessing mode is adopted as under the condition of the historical data.
Referring to fig. 7, in the present embodiment, step S21 includes:
s211: converting an original value into a tensor by using Pytorch as input to the value type value history data; for the string type value, firstly mapping the string value into a number corresponding to the code according to the mapping dictionary constructed in the step S1112, and then converting the number corresponding to the code into a tensor by using Pytorch as input; and converting the index value into a tensor using Pytorch;
s212: combining the numerical model value and the tensor converted by the character string model value, carrying out normalization operation on parameters, constructing a probability proxy model, fitting priori knowledge through Gaussian process regression to obtain approximate distribution of real problems, optimizing an acquisition function EI, and taking out experimental parameter combinations which enable the index improvement degree of the next round to be the largest as candidate samples.
In step S212, the selected collection function (acquisition function) is EI (Expected Improvement), and the EI function balances the exploration (extension) and utilization (extension) functions, and a compromise is taken between the exploration (extension) and the utilization (extension).
Referring to fig. 8, in the present embodiment, step S22 includes:
s221: the tensor composed of numerical values output by the Bayesian optimization algorithm is used as a column value, and the names of all experimental factors are used as column names to construct a data frame;
s222: outputting the nearest experimental factor value by utilizing the ordered list generated in the step S1111 aiming at the numerical value column;
s223: for the string type column, firstly, rounding operation is performed on each value by a round method in the Python scientific calculation package numpy, then, an integer obtained by the rounding operation is used as a key, the mapping dictionary constructed in the step S1112 is searched, and the string value corresponding to the key is taken out.
Specifically, in step S2 of this embodiment, the data after the initialization processing in step S1 is input into a round bayesian optimization algorithm, the round bayesian optimization algorithm constructs a probability proxy model, the prior knowledge is fitted through gaussian process regression to obtain the approximate distribution of the real problem, the collection function is optimized, the experimental parameter combination with the maximum improvement degree of the index of the next round is taken out as candidate samples, the candidate samples are tensor tensors composed of a plurality of groups of numerical values, the tensors are taken as column values, and the temperature, the pressure and the catalyst are taken as column names, so as to construct a data frame; for the temperature and pressure columns, the nearest experimental factor value can be output according to the ordered list generated in the step S1. For example, if the value output by the temperature column is 23.7, the processed value will be 20 as output; if the output value of the pressure column is 1.3, 1 is taken as output after being processed; for the catalyst column, firstly, rounding operation is carried out on each value by a round method in a Python scientific calculation package numpy, then the whole number is used as a key, the dictionary constructed in the step S1 is searched, and the character string value corresponding to the key is taken out. For example, the value output by the catalyst column is 0.213, the value is 0 after rounding, and the catalyst value corresponding to 0 is searched for as 'sample 1'.
And sequentially carrying out experiments on the recommended parameter combinations to obtain corresponding multi-index values, adding the corresponding multi-index values into a historical data set, inputting all the historical data as the next round of Bayesian optimization algorithm, realizing multi-round recommendation, and finally taking the multi-index values of a certain experimental parameter combination as the stopping condition of the round of Bayesian optimization algorithm when all the multi-index values of the certain experimental parameter combination reach an ideal state, so that experimenters can carry out mass production and manufacture according to the parameter combinations.
In summary, the present invention uses the names and the value types of the specified experimental factors and the parameter values of the specified experimental factors to perform permutation and combination to construct a full search space formed by mixing numerical values and character values for the current chemical experiment; designating the name of the index to be optimized, designating the expected optimization direction of the index to be optimized, and designating the weight corresponding to the index to be optimized. If an experimental scheme and a corresponding experimental index value exist, the experimental scheme and the corresponding experimental index value are directly used as a historical data set; if the experiment record does not exist, randomly generating a preset number of non-repeated experiment schemes from the search space to serve as initial samples, then carrying out experiments on the experiment schemes to obtain index values actually corresponding to each experiment scheme, and splicing the experiment schemes and the corresponding index values to obtain a historical data set. Converting an original value into a tensor by using Pytorch as input to the value type value history data; for the string type value, firstly mapping the string value into a number corresponding to the code according to the mapping dictionary constructed in the step S1112, and then converting the number corresponding to the code into a tensor by using Pytorch as input; and converting the index value into a tensor using Pytorch; combining the numerical model value and the tensor converted by the character string model value, carrying out normalization operation on parameters, constructing a probability proxy model, fitting priori knowledge through Gaussian process regression to obtain approximate distribution of real problems, optimizing an acquisition function EI, and taking out experimental parameter combinations which enable the index improvement degree of the next round to be the largest as candidate samples. The tensor composed of numerical values output by the Bayesian optimization algorithm is used as a column value, and the names of all experimental factors are used as column names to construct a data frame; outputting the nearest experimental factor value by utilizing the ordered list generated in the step S1111 aiming at the numerical value column; for the string type column, firstly, rounding operation is performed on each value by a round method in the Python scientific calculation package numpy, then, an integer obtained by the rounding operation is used as a key, the mapping dictionary constructed in the step S1112 is searched, and the string value corresponding to the key is taken out. According to the invention, as experimental data of multiple iterations are accumulated, the probability agent model can be more approximate to a real distribution curve, and at the moment, the optimized acquisition function can reasonably explore and utilize the residual area according to the curve, so that the target is continuously close to the set direction, and a scheme for combining experimental factor parameters, which enables multiple index values to reach better simultaneously, can be obtained after limited iterations, cumbersome and costly chemical experiment verification is omitted, the production cost can be effectively reduced, and the reasonable selection problem of experimental conditions can be effectively solved by related technicians.
It should be noted that the method of the embodiments of the present disclosure may be performed by a single device, such as a computer or a server. The method of the embodiment can also be applied to a distributed scene, and is completed by mutually matching a plurality of devices. In the case of such a distributed scenario, one of the devices may perform only one or more steps of the methods of embodiments of the present disclosure, the devices interacting with each other to accomplish the methods.
It should be noted that the foregoing describes some embodiments of the present disclosure. Other embodiments are within the scope of the following claims. In some cases, the actions or steps recited in the claims may be performed in a different order than in the embodiments described above and still achieve desirable results. In addition, the processes depicted in the accompanying figures do not necessarily require the particular order shown, or sequential order, to achieve desirable results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.
Example 2:
referring to fig. 9, embodiment 2 of the present invention further provides a chemical experiment multi-index optimization system, including:
The experimental information configuration module 11 is used for determining the to-be-optimized index and the experimental information of the experimental factors in the appointed aspect;
the data initialization module 12 is configured to use a differentiated data initialization method according to whether historical data exists in experimental information of experimental factors in a specified aspect;
the experimental combination scheme recommending module 21 is used for recommending a group of non-repeated experimental factor parameter combination schemes by taking the initialized historical data consisting of the numerical tensor as the input of a Bayesian optimization algorithm and through Gaussian process regression and preference comparison and selection;
the experiment combination scheme loop iteration module 22 is configured to perform experiments on the experiment factor parameter combination schemes recommended in the previous round, combine the experiment factor parameter combination schemes with the obtained corresponding index values, add the index values to the historical data set, take all the historical data as input of the bayesian optimization algorithm in the next round, recommend a set of experiment factor parameter combination schemes again for experiments, and perform loop iteration on the bayesian optimization algorithm for a plurality of rounds with each index value being in a preset range at the same time as a stop condition.
In this embodiment, the experimental information configuration module 11 includes:
the search space construction sub-module 111 is configured to perform permutation and combination on the current chemical experiment by using the name and the value type of the specified experimental factor and the parameter value of the specified experimental factor, and construct a full amount of search space formed by mixing the numerical value and the character value;
An index parameter setting submodule 112, configured to specify a name of the index to be optimized, specify a desired optimization direction of the index to be optimized, and specify a weight size corresponding to the index to be optimized;
the search space construction sub-module 111 includes:
a numerical value processing sub-module 1111, configured to sort all values of the experimental factors comprising numerical values to form an ordered list comprising a small number to a large number, and if more than one numerical experimental factor exists, form a nested list comprising a corresponding number of ordered lists;
a string type value processing sub-module 1112, configured to map and encode each of the different values of the experimental factors comprising string type values to form a mapping dictionary using the encoded number as a key and using the string as a value, and if more than one string type experimental factor exists, form a nested list comprising a corresponding number of mapping dictionaries;
the data initialization module 12 includes:
the experiment index value calling sub-module 121 is configured to directly use the existing experiment scheme and the corresponding experiment index value as a historical data set;
the experiment index value generating sub-module 122 is configured to randomly generate, if there is no experiment record, a preset number of non-repeated experiment schemes from the search space of the search space constructing sub-module 111 as initial samples, and then perform an experiment on the experiment schemes to obtain an index value actually corresponding to each experiment scheme, and splice the experiment schemes and the corresponding index values to obtain a historical data set.
In this embodiment, the experimental combination scheme recommendation module 21 includes:
a tensor generation sub-module 211 for converting the original value into a tensor using Pytorch as input to the numerical value history data; the mapping dictionary is also used for mapping the character string type value into numbers corresponding to codes according to the mapping dictionary constructed by the character string type value processing submodule, and then converting the numbers corresponding to the codes into tensor by using Pytorch as input; and converting the index value into a tensor using Pytorch;
the experimental parameter candidate sample generation sub-module 212 is configured to combine the tensor converted by the numerical model value and the character string model value, normalize the parameters, construct a probability proxy model, fit prior knowledge through gaussian process regression, obtain an approximate distribution of the real problem, optimize the collection function EI, and take out the experimental parameter combination that maximizes the improvement degree of the index of the next round as a candidate sample.
In this embodiment, the experimental combination scheme loop iteration module 22 includes:
the data frame construction sub-module 221 is configured to construct a data frame by taking a tensor composed of numerical values output by a bayesian optimization algorithm as a column value and taking names of all experimental factors as column names;
The experimental factor value output sub-module 222 is configured to output, for a numeric column, an experimental factor value closest to the ordered list generated by the numeric value processing sub-module 1111;
the rounding operation processing sub-module 223 is configured to perform rounding operation on each value by using a round method in the Python scientific calculation package numpy for the string type column, and then search the mapping dictionary constructed by the string type value processing sub-module 1112 by using the integer obtained by the rounding operation as a key, and take out the string value corresponding to the key.
It should be noted that, because the content of information interaction and execution process between the modules of the above system is based on the same concept as the method embodiment in the embodiment 1 of the present application, the technical effects brought by the content are the same as the method embodiment of the present application, and the specific content can be referred to the description in the foregoing illustrated method embodiment of the present application, which is not repeated herein.
Example 3:
embodiment 3 of the present application provides a non-transitory computer readable storage medium having stored therein program code for a chemical experiment multi-index optimization method, the program code comprising instructions for performing the chemical experiment multi-index optimization method of embodiment 1 or any possible implementation thereof.
Computer readable storage media can be any available media that can be accessed by a computer or data storage devices, such as servers, data centers, etc., that contain an integration of one or more available media. The usable medium may be a magnetic medium (e.g., floppy Disk, hard Disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium (e.g., solid State Disk, SSD), etc.
Example 4:
embodiment 4 of the present invention provides an electronic device, including: a memory and a processor;
the processor and the memory complete communication with each other through a bus; the memory stores program instructions executable by the processor to invoke a chemical experiment multi-index optimization method capable of performing embodiment 1 or any of its possible implementations.
Specifically, the processor may be implemented by hardware or software, and when implemented by hardware, the processor may be a logic circuit, an integrated circuit, or the like; when implemented in software, the processor may be a general-purpose processor, implemented by reading software code stored in a memory, which may be integrated in the processor, or may reside outside the processor, and which may reside separately.
In the above embodiments, it may be implemented in whole or in part by software, hardware, firmware, or any combination thereof. When implemented in software, may be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions. When loaded and executed on a computer, produces a flow or function in accordance with embodiments of the present invention, in whole or in part. The computer may be a general purpose computer, a special purpose computer, a network of computers, or other programmable devices. The computer instructions may be stored in or transmitted from one computer-readable storage medium to another, for example, by wired (e.g., coaxial cable, optical fiber, digital Subscriber Line (DSL)), or wireless (e.g., infrared, wireless, microwave, etc.).
It will be appreciated by those skilled in the art that the modules or steps of the invention described above may be implemented in a general purpose computing device, they may be concentrated on a single computing device, or distributed across a network of computing devices, they may alternatively be implemented in program code executable by computing devices, so that they may be stored in a memory device for execution by computing devices, and in some cases, the steps shown or described may be performed in a different order than that shown or described, or they may be separately fabricated into individual integrated circuit modules, or multiple modules or steps within them may be fabricated into a single integrated circuit module for implementation. Thus, the present invention is not limited to any specific combination of hardware and software.
While the invention has been described in detail in the foregoing general description and specific examples, it will be apparent to those skilled in the art that modifications and improvements can be made thereto. Accordingly, such modifications or improvements may be made without departing from the spirit of the invention and are intended to be within the scope of the invention as claimed.
Claims (10)
1. The multi-index optimization method for the chemical experiment is characterized by comprising the following steps of:
s1: experiment information configuration and data initialization:
s11: determining an index to be optimized and experimental information of experimental factors in a specified aspect;
s12: according to the existence of historical data of experimental information of experimental factors in the appointed aspect, adopting a differentiated data initialization method;
s2: reading in the initialized experimental information data, and carrying out iterative recommendation:
s21: taking the initialized historical data consisting of the numerical tensors as the input of a Bayesian optimization algorithm, and recommending a group of non-repeated experimental factor parameter combination schemes through Gaussian process regression and preference comparison selection;
s22: and respectively carrying out experiments on the recommended experimental factor parameter combination schemes of the previous round, merging the recommended experimental factor parameter combination schemes with the obtained corresponding index values, adding the merged experimental factor parameter combination schemes to a historical data set, taking all historical data as the input of a Bayesian optimization algorithm of the next round, recommending a set of experimental factor parameter combination schemes for experiments again, and carrying out a plurality of rounds of loop iteration on the Bayesian optimization algorithm by taking the condition that all the index values are simultaneously in a preset range as stop conditions.
2. The method for optimizing multiple indices of chemical experiments according to claim 1, wherein step S11 comprises:
s111: the method comprises the steps of performing permutation and combination on a current chemical experiment by using the names and the value types of experimental factors in the appointed aspect and the parameter values of the experimental factors in the appointed aspect to construct a full search space formed by mixing numerical type values and character type values;
s112: designating the name of the index to be optimized, designating the expected optimization direction of the index to be optimized, and designating the weight corresponding to the index to be optimized.
3. The method for optimizing multiple indices of chemical experiments according to claim 2, wherein step S111 comprises:
s1111: sequencing all values of the experimental factors to form an ordered list from small to large, and forming a nested list consisting of a corresponding number of ordered lists if more than one numerical experimental factor exists;
s1112: and mapping and encoding each different value to form a mapping dictionary taking the encoded number as a key and taking the character string as a value, and if more than one character string type experimental factors exist, forming a nested list consisting of a corresponding number of mapping dictionaries.
4. A method for optimizing multiple indices of a chemical experiment according to claim 3, wherein step S12 comprises:
s121: if an experimental scheme and a corresponding experimental index value exist, the experimental scheme and the corresponding experimental index value are directly used as a historical data set;
s122: if no experiment record exists, randomly generating a preset number of non-repeated experiment schemes from the search space in the step S111 to serve as initial samples, then carrying out experiments on the experiment schemes to obtain index values actually corresponding to each experiment scheme, and splicing the experiment schemes and the corresponding index values to obtain a historical data set.
5. The method for optimizing multiple indices of chemical experiments according to claim 4, wherein step S21 comprises:
s211: converting an original value into a tensor by using Pytorch as input to the value type value history data; for the string type value, firstly mapping the string value into a number corresponding to the code according to the mapping dictionary constructed in the step S1112, and then converting the number corresponding to the code into a tensor by using Pytorch as input; and converting the index value into a tensor using Pytorch;
s212: combining the numerical model value and the tensor converted by the character string model value, carrying out normalization operation on parameters, constructing a probability proxy model, fitting priori knowledge through Gaussian process regression to obtain approximate distribution of real problems, optimizing an acquisition function EI, and taking out experimental parameter combinations which enable the index improvement degree of the next round to be the largest as candidate samples.
6. The method for optimizing multiple indices of chemical experiments according to claim 5, wherein step S22 comprises:
s221: the tensor composed of numerical values output by the Bayesian optimization algorithm is used as a column value, and the names of all experimental factors are used as column names to construct a data frame;
s222: outputting the nearest experimental factor value by utilizing the ordered list generated in the step S1111 aiming at the numerical value column;
s223: for the string type column, firstly, rounding operation is performed on each value by a round method in the Python scientific calculation package numpy, then, an integer obtained by the rounding operation is used as a key, the mapping dictionary constructed in the step S1112 is searched, and the string value corresponding to the key is taken out.
7. A chemical experiment multi-index optimization system, comprising:
the experimental information configuration module is used for determining the to-be-optimized index and the experimental information of the experimental factors in the appointed aspect;
the data initialization module is used for adopting a differential data initialization method according to whether historical data exists in experimental information of experimental factors in the appointed aspect;
the experimental combination scheme recommending module is used for recommending a group of non-repeated experimental factor parameter combination schemes by taking the initialized historical data consisting of the numerical tensor as the input of a Bayesian optimization algorithm and through Gaussian process regression and preference comparison and selection;
And the experiment combination scheme loop iteration module is used for respectively carrying out experiments on the experiment factor parameter combination schemes recommended in the previous round, merging the experiment factor parameter combination schemes with the obtained corresponding index values, adding the merged index values to a historical data set, taking all the historical data as the input of the Bayesian optimization algorithm in the next round, recommending a group of experiment factor parameter combination schemes again for the experiments, and carrying out loop iteration on the Bayesian optimization algorithm for a plurality of rounds by taking the condition that all the index values are simultaneously in a preset range.
8. The chemical experiment multi-index optimization system of claim 7, wherein the experiment information configuration module comprises:
the search space construction submodule is used for carrying out permutation and combination on the current chemical experiment by utilizing the names and the value types of the experimental factors in the appointed aspect and the parameter values of the experimental factors in the appointed aspect to construct a full amount of search space formed by mixing the numerical type values and the character type values;
the index parameter setting submodule is used for designating the name of the index to be optimized, designating the expected optimization direction of the index to be optimized and designating the weight corresponding to the index to be optimized;
the search space construction submodule includes:
The numerical value processing submodule is used for sequencing all values of experimental factors consisting of numerical values to form an ordered list from small to large, and if more than one numerical experimental factor exists, a nested list consisting of the ordered list with corresponding number is formed;
the character string type value processing sub-module is used for mapping and encoding each different value of experimental factors consisting of character string type values to form a mapping dictionary taking an encoded number as a key and taking a character string as a value, and if more than one character string type experimental factor exists, a nested list consisting of a corresponding number of mapping dictionaries is formed;
the data initialization module comprises:
the experimental index value calling sub-module is used for directly using the experimental index value as a historical data set if an experimental scheme and a corresponding experimental index value exist;
and the experiment index value generation sub-module is used for randomly generating a preset number of non-repeated experiment schemes in the search space of the search space construction sub-module as initial samples if no experiment record exists, carrying out experiments on the experiment schemes to obtain index values actually corresponding to each experiment scheme, and splicing the experiment schemes and the corresponding index values to obtain a historical data set.
9. The chemical experiment multi-index optimization system of claim 8, wherein the experiment combination scheme recommendation module comprises:
the tensor generation submodule is used for converting an original value into a tensor by using Pytorch as input to the numerical value historical data; the mapping dictionary is also used for mapping the character string type value into numbers corresponding to codes according to the mapping dictionary constructed by the character string type value processing submodule, and then converting the numbers corresponding to the codes into tensor by using Pytorch as input; and converting the index value into a tensor using Pytorch;
the experimental parameter candidate sample generation sub-module is used for combining the tensor converted by the numerical model value and the character string model value, carrying out normalization operation on the parameters, constructing a probability agent model, fitting priori knowledge through Gaussian process regression to obtain approximate distribution of real problems, optimizing an acquisition function EI, and taking out the experimental parameter combination which enables the index improvement degree of the next round to be the largest as a candidate sample.
10. The chemical experiment multi-index optimization system of claim 9, wherein the experiment combination scheme loop iteration module comprises:
The data frame construction submodule is used for constructing a data frame by taking a tensor consisting of numerical values output by a Bayesian optimization algorithm as a column value and taking the names of all experimental factors as column names;
the experimental factor value output sub-module is used for outputting the experimental factor value closest to the ordered list generated by the numerical value processing sub-module aiming at the numerical value column;
the rounding operation processing sub-module is used for carrying out rounding operation on each value by a round method in the Python scientific calculation packet numpy aiming at the character string type column, and searching a mapping dictionary constructed by the character string type value processing sub-module by taking an integer obtained by the rounding operation as a key to take out a character string value corresponding to the key.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311218216.0A CN116959629B (en) | 2023-09-21 | 2023-09-21 | Multi-index optimization method and system for chemical experiment, storage medium and electronic equipment |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311218216.0A CN116959629B (en) | 2023-09-21 | 2023-09-21 | Multi-index optimization method and system for chemical experiment, storage medium and electronic equipment |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN116959629A true CN116959629A (en) | 2023-10-27 |
| CN116959629B CN116959629B (en) | 2023-12-29 |
Family
ID=88462489
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202311218216.0A Active CN116959629B (en) | 2023-09-21 | 2023-09-21 | Multi-index optimization method and system for chemical experiment, storage medium and electronic equipment |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116959629B (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117649898A (en) * | 2024-01-30 | 2024-03-05 | 烟台国工智能科技有限公司 | Liquid crystal material formula analysis method and device based on data mining |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112101630A (en) * | 2020-08-19 | 2020-12-18 | 江苏师范大学 | Multi-target optimization method for injection molding process parameters of thin-wall plastic part |
| WO2022167821A1 (en) * | 2021-02-08 | 2022-08-11 | Exscientia Limited | Drug optimisation by active learning |
| CN115129982A (en) * | 2022-06-22 | 2022-09-30 | 电计科技发展(上海)有限公司 | Experiment parameter recommendation method, device, terminal and medium based on improved Bayesian optimization |
| CN115376621A (en) * | 2022-08-16 | 2022-11-22 | 北京晶泰科技有限公司 | Method and device for assisting optimization process and electronic equipment |
| CN116110505A (en) * | 2022-11-30 | 2023-05-12 | 浙江工业大学 | Flow chemistry process optimization method based on multi-objective Bayesian optimization |
| US20230281363A1 (en) * | 2022-03-03 | 2023-09-07 | International Business Machines Corporation | Optimal materials and devices design using artificial intelligence |
| CN116764571A (en) * | 2023-06-06 | 2023-09-19 | 浙江工业大学 | Photocatalytic CO based on machine learning 2 Reduction reaction condition optimization method |
-
2023
- 2023-09-21 CN CN202311218216.0A patent/CN116959629B/en active Active
Patent Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112101630A (en) * | 2020-08-19 | 2020-12-18 | 江苏师范大学 | Multi-target optimization method for injection molding process parameters of thin-wall plastic part |
| WO2022167821A1 (en) * | 2021-02-08 | 2022-08-11 | Exscientia Limited | Drug optimisation by active learning |
| CN116601715A (en) * | 2021-02-08 | 2023-08-15 | 艾克斯赛安西娅人工智能有限公司 | Drug optimization through active learning |
| US20230281363A1 (en) * | 2022-03-03 | 2023-09-07 | International Business Machines Corporation | Optimal materials and devices design using artificial intelligence |
| CN115129982A (en) * | 2022-06-22 | 2022-09-30 | 电计科技发展(上海)有限公司 | Experiment parameter recommendation method, device, terminal and medium based on improved Bayesian optimization |
| CN115376621A (en) * | 2022-08-16 | 2022-11-22 | 北京晶泰科技有限公司 | Method and device for assisting optimization process and electronic equipment |
| CN116110505A (en) * | 2022-11-30 | 2023-05-12 | 浙江工业大学 | Flow chemistry process optimization method based on multi-objective Bayesian optimization |
| CN116764571A (en) * | 2023-06-06 | 2023-09-19 | 浙江工业大学 | Photocatalytic CO based on machine learning 2 Reduction reaction condition optimization method |
Non-Patent Citations (1)
| Title |
|---|
| 崔佳旭等: "贝叶斯优化方法和应用综述", 软件学报, vol. 29, no. 10, pages 3068 - 3090 * |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117649898A (en) * | 2024-01-30 | 2024-03-05 | 烟台国工智能科技有限公司 | Liquid crystal material formula analysis method and device based on data mining |
| CN117649898B (en) * | 2024-01-30 | 2024-05-03 | 烟台国工智能科技有限公司 | Liquid crystal material formula analysis method and device based on data mining |
Also Published As
| Publication number | Publication date |
|---|---|
| CN116959629B (en) | 2023-12-29 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Kouzes et al. | The changing paradigm of data-intensive computing | |
| CN116959629B (en) | Multi-index optimization method and system for chemical experiment, storage medium and electronic equipment | |
| CN113535984A (en) | Attention mechanism-based knowledge graph relation prediction method and device | |
| US20190327274A1 (en) | Positionally-encoded string representations, including their use in machine learning | |
| Cao et al. | A bottom-up DAG structure extraction model for math word problems | |
| Farahat et al. | Distributed column subset selection on mapreduce | |
| RU2693324C2 (en) | Method and a server for converting a categorical factor value into its numerical representation | |
| CN112994701B (en) | Data compression method, device, electronic equipment and computer readable medium | |
| CN103336791A (en) | Hadoop-based fast rough set attribute reduction method | |
| Huang et al. | Position-enhanced and time-aware graph convolutional network for sequential recommendations | |
| CN116959613B (en) | Compound inverse synthesis method and device based on quantum mechanical descriptor information | |
| Soltaniyeh et al. | An accelerator for sparse convolutional neural networks leveraging systolic general matrix-matrix multiplication | |
| Xu et al. | Design and implementation of the modified signed digit multiplication routine on a ternary optical computer | |
| Hu et al. | On Computational Limits of Modern Hopfield Models: A Fine-Grained Complexity Analysis | |
| CN114389843A (en) | Network abnormal intrusion detection system and method based on variational self-encoder | |
| US9665538B2 (en) | Solving satisfiability problems through search | |
| CN113076545A (en) | Deep learning-based kernel fuzzy test sequence generation method | |
| Qiu et al. | Efficient document retrieval by end-to-end refining and quantizing BERT embedding with contrastive product quantization | |
| CN116738081A (en) | Front-end component binding method, device and storage medium | |
| WO2023078009A1 (en) | Model weight acquisition method and related system | |
| Niu et al. | Fair: Flow type-aware pre-training of compiler intermediate representations | |
| Goto et al. | Fast q-gram mining on SLP compressed strings | |
| Wang et al. | Robust Recommendation with Adversarial Gaussian Data Augmentation | |
| Ruberto et al. | A semantic genetic programming framework based on dynamic targets | |
| Song et al. | Large scale network embedding: A separable approach |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |