CN112633517B - Training method of machine learning model, computer equipment and storage medium - Google Patents
Training method of machine learning model, computer equipment and storage medium Download PDFInfo
- Publication number
- CN112633517B CN112633517B CN202011591857.7A CN202011591857A CN112633517B CN 112633517 B CN112633517 B CN 112633517B CN 202011591857 A CN202011591857 A CN 202011591857A CN 112633517 B CN112633517 B CN 112633517B
- Authority
- CN
- China
- Prior art keywords
- sample data
- loss function
- sample
- function value
- twin
- 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.)
- Active
Links
- 238000012549 training Methods 0.000 title claims abstract description 79
- 238000010801 machine learning Methods 0.000 title claims abstract description 69
- 238000000034 method Methods 0.000 title claims abstract description 62
- 230000006870 function Effects 0.000 claims abstract description 265
- 238000004364 calculation method Methods 0.000 claims abstract description 76
- 238000000605 extraction Methods 0.000 claims description 35
- 238000004590 computer program Methods 0.000 claims description 2
- 238000011161 development Methods 0.000 description 8
- 238000004422 calculation algorithm Methods 0.000 description 7
- 238000010586 diagram Methods 0.000 description 6
- 238000012937 correction Methods 0.000 description 5
- 230000006978 adaptation Effects 0.000 description 4
- 238000013145 classification model Methods 0.000 description 4
- 238000001514 detection method Methods 0.000 description 3
- 230000000694 effects Effects 0.000 description 3
- 238000012545 processing Methods 0.000 description 3
- 238000013473 artificial intelligence Methods 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 230000009466 transformation Effects 0.000 description 2
- 238000013459 approach Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 238000007405 data analysis Methods 0.000 description 1
- 238000013500 data storage Methods 0.000 description 1
- 230000001419 dependent effect Effects 0.000 description 1
- 238000007636 ensemble learning method Methods 0.000 description 1
- 238000005286 illumination Methods 0.000 description 1
- 238000010295 mobile communication Methods 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 230000008707 rearrangement Effects 0.000 description 1
- 238000012827 research and development Methods 0.000 description 1
- 238000005070 sampling Methods 0.000 description 1
- 238000012163 sequencing technique Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Software Systems (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Medical Informatics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Physics & Mathematics (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Artificial Intelligence (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a training method of a machine learning model, computer equipment and a storage medium, wherein the method comprises the following steps: acquiring a plurality of first sample data and second sample data from a plurality of sample domains; inputting the first sample data into a target model to obtain a plurality of first loss function values, and performing backward calculation to obtain a plurality of discrimination network parameters; inputting the second sample data into a target model for loading each discrimination network parameter to obtain a plurality of second loss function values; and according to the first loss function value and the second loss function value, calculating to obtain a total loss function value, carrying out backward calculation according to the total loss function value, and returning to execute the operation of acquiring a plurality of first sample data and second sample data from a plurality of sample domains until the end training condition is met. The technical scheme of the embodiment of the invention can realize the automation of the training process of the machine learning model, and effectively improve the robustness of the machine learning model in different scenes.
Description
Technical Field
The embodiment of the invention relates to the technical field of artificial intelligence, in particular to a training method of a machine learning model, computer equipment and a storage medium.
Background
With research and development of data analysis technology and artificial intelligence technology, people pay more attention to the robustness of a machine learning model in different application scenes. The machine learning model with high robustness can maintain certain performances in a new scene, and has high effectiveness.
In order to improve the robustness of a machine learning model, the existing method is to introduce regularization or dropout and other algorithms in the model training process; or fine tuning the machine learning model, namely training the training sample of the new application scene on the basis of the pre-training model.
Although the regularization or dropout and other algorithms are introduced to improve the robustness of the machine learning model to a certain extent, the stability of the machine learning model is easily reduced, and the final effect of the model is often dependent on experience of a developer and a large amount of trial and error; by fine tuning the machine learning model, the final effect of the model is generally less good than the model retraining effect, and whichever approach described above requires the developer to do the adaptation of the model to the various scenarios and maintain numerous model versions.
Disclosure of Invention
The embodiment of the invention provides a training method, computer equipment and storage medium for a machine learning model, which can realize the automation of the training process of the machine learning model and effectively improve the robustness of the machine learning model in different scenes.
In a first aspect, an embodiment of the present invention provides a training method for a machine learning model, including:
acquiring a plurality of first sample data and second sample data from a plurality of sample domains, each sample domain including a plurality of sample data corresponding to the same scene;
inputting each first sample data into a target model, performing forward calculation on each first sample data by using a feature extraction network and a discrimination network in the target model respectively to obtain a plurality of first loss function values, and performing backward calculation to obtain a plurality of discrimination network parameters;
inputting each second sample data into a target model for loading each discrimination network parameter, and performing forward calculation on each second sample data to obtain a plurality of second loss function values;
and according to the first loss function value and the second loss function value, calculating to obtain a total loss function value, carrying out backward calculation according to the total loss function value, and returning to execute the operation of acquiring a plurality of first sample data and second sample data from a plurality of sample domains until the end training condition is met.
In a second aspect, embodiments of the present invention also provide a computer device comprising a processor and a memory for storing instructions that, when executed, cause the processor to:
acquiring a plurality of first sample data and second sample data from a plurality of sample domains, each sample domain including a plurality of sample data corresponding to the same scene;
inputting each first sample data into a target model, performing forward calculation on each first sample data by using a feature extraction network and a discrimination network in the target model respectively to obtain a plurality of first loss function values, and performing backward calculation to obtain a plurality of discrimination network parameters;
inputting each second sample data into a target model for loading each discrimination network parameter, and performing forward calculation on each second sample data to obtain a plurality of second loss function values;
and according to the first loss function value and the second loss function value, calculating to obtain a total loss function value, carrying out backward calculation according to the total loss function value, and returning to execute the operation of acquiring a plurality of first sample data and second sample data from a plurality of sample domains until the end training condition is met.
In a third aspect, an embodiment of the present invention further provides a storage medium, where the storage medium is configured to store instructions for performing:
acquiring a plurality of first sample data and second sample data from a plurality of sample domains, each sample domain including a plurality of sample data corresponding to the same scene;
inputting each first sample data into a target model, performing forward calculation on each first sample data by using a feature extraction network and a discrimination network in the target model respectively to obtain a plurality of first loss function values, and performing backward calculation to obtain a plurality of discrimination network parameters;
inputting each second sample data into a target model for loading each discrimination network parameter, and performing forward calculation on each second sample data to obtain a plurality of second loss function values;
and according to the first loss function value and the second loss function value, calculating to obtain a total loss function value, carrying out backward calculation according to the total loss function value, and returning to execute the operation of acquiring a plurality of first sample data and second sample data from a plurality of sample domains until the end training condition is met.
According to the embodiment of the invention, a plurality of first sample data and second sample data are acquired from a plurality of sample domains, each first sample data is input into a target model, a feature extraction network and a discrimination network in the target model are respectively used for carrying out forward calculation on each first sample data to obtain a plurality of first loss function values, and backward calculation is carried out to obtain a plurality of discrimination network parameters; inputting each second sample data into a target model corresponding to each discrimination network parameter, and performing forward calculation on each second sample data to obtain a plurality of second loss function values; according to the first loss function value and the second loss function value, the total loss function value is obtained through calculation, backward calculation is carried out according to the total loss function value, the operation of obtaining a plurality of first sample data and second sample data from a plurality of sample domains is carried out in a return mode until the end training condition is met, automation of the machine learning model training process can be achieved, robustness of the machine learning model in different scenes is effectively improved, and development cost of the machine learning model is saved.
Drawings
FIG. 1 is a flowchart of a training method of a machine learning model according to an embodiment of the present invention;
FIG. 2a is a flowchart of a training method of a machine learning model according to a second embodiment of the present invention;
FIG. 2b is a schematic diagram of a training method of a machine learning model according to a second embodiment of the present invention;
fig. 3 is a schematic structural diagram of a training device for a machine learning model according to a third embodiment of the present invention;
fig. 4 is a schematic structural diagram of a computer device according to a fourth embodiment of the present invention.
Detailed Description
The invention is described in further detail below with reference to the drawings and examples. It is to be understood that the specific embodiments described herein are merely illustrative of the invention and are not limiting thereof. It should be further noted that, for convenience of description, only some, but not all of the structures related to the present invention are shown in the drawings.
Before discussing exemplary embodiments in more detail, it should be mentioned that some exemplary embodiments are described as processes or methods depicted as flowcharts. Although a flowchart depicts operations (or steps) as a sequential process, many of the operations can be performed in parallel, concurrently, or at the same time. Furthermore, the order of the operations may be rearranged. The process may be terminated when its operations are completed, but may have additional steps not included in the figures. The processes may correspond to methods, functions, procedures, subroutines, and the like.
The term "target model" as used herein refers to a machine learning model to be trained that is the result of the machine learning algorithm's yield, which can be considered to output a certain result given an input. For example, when the machine learning model is a specific classification model, the output result may be category information corresponding to the input data.
The term "feature extraction network" is used herein to extract features corresponding to input data, the "discrimination network" is used to discriminate the input data according to the features of the input data, and the "feature extraction network" and the "discrimination network" together form the "object model".
The main inventive concept of the embodiments of the present invention will be briefly described for easy understanding.
The robustness of the machine learning model in different application scenarios is particularly important for the model in practical engineering applications. In the prior art, in order to improve the robustness of a machine learning model, algorithms such as regularization or dropout are generally introduced; or fine tuning the machine learning model, namely training the training sample of the new application scene on the basis of the pre-training model. However, the existing machine learning model training method generally requires a developer to perform model adaptation work for various scenes, and maintains numerous model versions, so that the model development cost is high.
The inventor aims at the problem that the adaptation work of the machine learning model needs to be completed manually in the prior art, considers whether the automatic training of the machine learning model can be realized through one method, and can improve the robustness of the machine learning model in different scenes, so that the effectiveness of the machine learning model can be improved, and the development cost of the model can be saved.
Based on the above-mentioned thinking, the inventors creatively propose that, from a plurality of sample domains, a plurality of first sample data and second sample data are obtained, each first sample data is input into a target model, forward computation is performed on each first sample data by using a feature extraction network and a discrimination network in the target model, a plurality of first loss function values are obtained, backward computation is performed, and a plurality of discrimination network parameters are obtained; inputting each second sample data into a target model loaded with each discrimination network parameter, and performing forward calculation on each second sample data to obtain a plurality of second loss function values; according to the first loss function value and the second loss function value, the total loss function value is obtained through calculation, backward calculation is carried out according to the total loss function value, the operation of obtaining a plurality of first sample data and second sample data from a plurality of sample domains is carried out in a return mode until the end training condition is met, automation of the machine learning model training process can be achieved, robustness of the machine learning model in different scenes is effectively improved, and development cost of the machine learning model is saved.
Example 1
Fig. 1 is a flowchart of a training method of a machine learning model according to an embodiment of the present invention. The embodiment of the invention can be suitable for the condition of training the machine learning model, the method can be executed by the training device of the machine learning model, and the device can be realized in a software and/or hardware mode and can be generally integrated in computer equipment. As shown in fig. 1, the method in the embodiment of the present invention specifically includes:
step 110, acquiring a plurality of first sample data and second sample data from a plurality of sample domains, wherein each sample domain comprises a plurality of sample data corresponding to the same scene.
In this embodiment, the sample fields are a set including a plurality of samples, each sample field corresponds to a specific scene, each sample field includes a plurality of sample data corresponding to the scene, and different sample fields correspond to different scenes. The scene may specifically be source information of sample data in a sample domain, for example, a device type of collecting a sample, a background condition of generating the sample (such as a lighting condition when generating a picture sample, an indoor and outdoor scene, and the like).
In this step, alternatively, the respective corresponding sample data may be acquired in different sample domains, with the sample data corresponding to several of the sample domains being regarded as first sample data, and the sample data corresponding to the other sample domains being regarded as second sample data. The first sample data and the second sample data are used together for performing iterative training on a target model.
And 120, inputting the first sample data into a target model, performing forward computation on the first sample data by using a feature extraction network and a discrimination network in the target model to obtain a plurality of first loss function values, and performing backward computation to obtain a plurality of discrimination network parameters.
The target model is a machine learning model to be trained, and specifically, the target model may be a classification model, a regression model, a data clustering model, and the like. The target model comprises a feature extraction network and a discrimination network.
In this embodiment, the training process of the machine learning model can be divided into two phases: the method comprises an exploration stage and a correction stage. The first sample data are sample data participating in the exploration phase, and the second sample data are sample data participating in the correction phase.
The exploring stage is to input each first sample data into the target model, forward calculate each first sample data through a feature extraction network and a discrimination network in the target model to obtain a first loss function value corresponding to each first sample data, and then backward calculate each first sample data through the discrimination network in the target model to obtain a discrimination network parameter corresponding to the first loss function value.
The first sample data can be input into the target model in parallel, so that training time of the target model can be reduced, and training efficiency of the target model can be improved.
And 130, inputting each second sample data into a target model loaded with each discrimination network parameter, and performing forward calculation on each second sample data to obtain a plurality of second loss function values.
In this embodiment, the correction stage refers to that after the discrimination network parameters are calculated, the objective model is made to load each discrimination network parameter in turn, then each second sample data is input into the objective model loaded with different discrimination network parameters, and each second loss function value is obtained by extracting the network from the characteristics in the objective model and performing forward calculation on each second sample data by the discrimination network. It should be noted that, unlike the ensemble learning method, only one target model exists in this embodiment all the time.
And 140, calculating to obtain a total loss function value according to each first loss function value and each second loss function value, performing backward calculation according to the total loss function value, and returning to execute the operation of acquiring a plurality of first sample data and second sample data from a plurality of sample domains until the end training condition is met.
In this step, the first loss function values and the second loss function values may be tensed to obtain total loss function values, or the first loss function values and the second loss function values may be tensed to obtain total loss function values according to a preset weight ratio.
In this embodiment, after the total loss function value is calculated, the total loss function value may be calculated backward by the target model, parameters of the target model (for example, weight and bias of the target model) are adjusted according to the calculation result, then the first sample data and the second sample data are continuously acquired from the plurality of sample domains, and the total loss function value is calculated again until the target model meets the end training condition. When the total loss function value is smaller than a preset threshold value, it may be determined that the target model converges, that is, the target model satisfies the training ending condition.
In the embodiment, the first sample data and the second sample data corresponding to different scenes are obtained, so that the training of the target model by the multi-scene sample data can be realized, the robustness of the target model in different scenes can be improved, a developer does not need to do the adaptation of the target model to various scenes, the automation of the training process of the target model can be realized, and the development cost of the target model is saved; and secondly, carrying out forward computation on the first sample data through the feature extraction network and the discrimination network to obtain a first loss function value and discrimination network parameters, then carrying out forward computation on the second sample data under the condition that the target model loads each discrimination network parameter to obtain each second loss function value, and calculating to obtain a total loss function value according to the first loss function value and each second loss function value, wherein the feature extraction network and the discrimination network are combined to resist learning results, and parameter adjustment is effectively carried out on the target model, so that training efficiency of the target model can be improved.
According to the embodiment of the invention, a plurality of first sample data and second sample data are acquired from a plurality of sample domains, each first sample data is input into a target model, a feature extraction network and a discrimination network in the target model are respectively used for carrying out forward calculation on each first sample data to obtain a plurality of first loss function values, and backward calculation is carried out to obtain a plurality of discrimination network parameters; inputting each second sample data into a target model loaded with each discrimination network parameter, and performing forward calculation on each second sample data to obtain a plurality of second loss function values; according to the first loss function value and the second loss function value, the total loss function value is obtained through calculation, backward calculation is carried out according to the total loss function value, the operation of obtaining a plurality of first sample data and second sample data from a plurality of sample domains is carried out in a return mode until the end training condition is met, automation of the machine learning model training process can be achieved, robustness of the machine learning model in different scenes is effectively improved, and development cost of the machine learning model is saved.
Example two
The present embodiment is a further refinement of the first embodiment, and the same or corresponding terms as those of the first embodiment are explained, and the description of the present embodiment is omitted. Fig. 2a is a flowchart of a training method of a machine learning model according to a second embodiment of the present invention, in this embodiment, the technical solution of this embodiment may be combined with one or more methods in the solutions of the foregoing embodiments, and in this embodiment, as shown in fig. 2a, the method provided by the embodiment of the present invention may further include:
Step 210, acquiring a plurality of sample domains, and acquiring corresponding batch sample data from each sample domain.
Wherein each sample domain includes a plurality of sample data corresponding to the same scene.
In one implementation of the embodiment of the present invention, obtaining a plurality of sample fields includes: acquiring a plurality of sample data, and dividing each sample data into a plurality of sample sets according to a preset classification rule; and adjusting the proportion of the sample data of different types in each sample set and the proportion of the sample data among each sample set to obtain a plurality of sample domains.
In this embodiment, the classification rule may be constructed according to the data source attribute corresponding to the sample data and the tag attribute of the sample data. The data source attribute may be a type of a device that generates the sample data, a size of the sample data, a background condition that forms the sample data, a field to which the sample data belongs, and the like. The tag attribute may be tag information carried by the sample data, for example, when the sample data is face picture data, the tag attribute may be information of a corresponding age, gender, race, and the like.
In a specific embodiment, assuming that the sample data is picture data, the background conditions for forming the sample data may be illumination conditions, indoor and outdoor scenes, and the like; the sample data is assumed to be a photograph containing a helmet, and the field to which the sample data belongs may be a chemical plant or a construction site.
In this embodiment, after a plurality of sample data are obtained, each sample data is first allocated to a corresponding sample set according to the established classification rule, where each sample set corresponds to a specific data source attribute or tag attribute. In order to ensure that sample data in each sample set reach equilibrium, a preset sample equilibrium method can be used, and the proportion of different types of sample data in each sample set and the proportion of sample data among each sample set are adjusted according to the requirement of super parameters, so that a plurality of sample domains are obtained. The sample equalization method may include a sampling method and a data enhancement method.
In a specific embodiment, assuming that the target model is a classification model, and a sample set includes five sample pictures of type a and ten sample pictures of type B, a data enhancement method may be used to perform geometric transformation (such as flipping, rotation, clipping, deformation, scaling, etc.) on the sample pictures of type a, so that the number of sample pictures of type a is equal to the number of sample pictures of type B. Assuming that a total of six sample pictures are included in one sample set and a total of ten sample pictures are included in the other sample set, the geometric transformation can be performed on the pictures in the previous sample set by adopting the data enhancement method, so that the number of sample pictures in the two sample sets is equal.
Therefore, through establishing a classification rule, distributing each sample data to a corresponding sample set, and ensuring that samples included in each sample set correspond to the same scene; the sample data in each sample set can be balanced by adjusting the proportion of different types of sample data in each sample set and the proportion of sample data among each sample set, so that the target model can be fairly trained by the multi-scene samples, and the training efficiency of the target model and the effectiveness of the training result of the target model can be improved.
In one implementation of the embodiment of the present invention, obtaining corresponding batch sample data from each sample domain includes: and acquiring batch sample data matched with the batch threshold value from each sample domain according to the preset batch threshold value through a data loader corresponding to each sample domain.
Each sample domain corresponds to one data loader, and batch sample data are acquired through the data loader, so that automation of data acquisition can be realized, and further training efficiency of a target model is improved.
Step 220, randomly selecting the batch sample data corresponding to the plurality of sample domains from all the batch sample data as first sample data, and taking the batch sample data corresponding to the rest of the plurality of sample domains as second sample data.
In this embodiment, among all the batch sample data, the batch sample data corresponding to the plurality of sample fields may be randomly selected, each of the batch sample data is taken as first sample data, and then the remaining batch sample data corresponding to each sample field is taken as second sample data.
The number of the first sample data may be equal to the number of the second sample data, or when the number of the first sample data is not equal to the number of the second sample data, a difference between the number of the first sample data and the number of the second sample data needs to be smaller than a preset threshold. Therefore, the situation that the quantity of the first sample data is too large compared with the quantity of the second sample data can be avoided, the situation that the quantity of the sample data participating in the exploration stage and the correction stage is similar can be guaranteed, and then the goal model can be fairly trained by the multi-scene sample data, so that the training efficiency of the goal model is improved, and the effectiveness of the training result of the goal model is improved.
Step 230, inputting each first sample data into the target model, performing forward computation on each first sample data by using the feature extraction network and the discrimination network in the target model to obtain a plurality of first loss function values, and performing backward computation to obtain a plurality of discrimination network parameters.
The first sample data can be input into the target model in parallel, so that training time of the target model can be reduced, and training efficiency of the target model can be improved.
In one implementation manner of the embodiment of the present invention, after obtaining the plurality of discrimination network parameters, the method further includes: and acquiring the characteristics corresponding to the first sample data output by the characteristic extraction network in the target model, and calculating a first twin loss function value through a twin network loss function according to the characteristics corresponding to the first sample data.
Step 240, inputting each second sample data into the target model loaded with each discrimination network parameter, and performing forward calculation on each second sample data to obtain a plurality of second loss function values.
In this embodiment, optionally, after obtaining the plurality of second loss function values, the method further includes: and acquiring the characteristics corresponding to the second sample data output by the characteristic extraction network in the target model, and calculating a second twin loss function value through a twin network loss function according to the characteristics corresponding to the second sample data.
Step 250, calculating to obtain a total loss function value according to each first loss function value and each second loss function value, performing backward calculation according to the total loss function value, and returning to execute the operation of obtaining a plurality of first sample data and second sample data from a plurality of sample domains until the end training condition is met.
In this embodiment, optionally, the calculating a total loss function value according to each first loss function value and each second loss function value, and performing a backward calculation according to the total loss function value includes: and calculating to obtain a total loss function value according to the first loss function value, the second loss function value, the first twin loss function value and the second twin loss function value, performing backward calculation according to the total loss function value, and updating the parameters of the target model according to the calculation result.
Therefore, the convergence speed of the target model can be increased, and the training efficiency of the target model can be improved by calculating the first twin loss function value and the second twin loss function value and combining the first twin loss function value and the second twin loss function value to calculate the total loss function value.
In a specific embodiment, the calculating the total loss function value from each of the first loss function value, the second loss function value, the first twin loss function value, and the second twin loss function value includes: tensor adding the first loss function value and the second loss function value to obtain a first function value; tensor adding the first twin loss function value and the second twin loss function value to obtain a second function value; and calculating the total loss function value according to the first function value, the second function value and the super parameter corresponding to the target model.
The second function value and the super parameter corresponding to the target model can be multiplied to obtain a third function value, and finally, the third function value and the first function value are tensor added to obtain the total loss function value.
According to the embodiment of the invention, a plurality of sample domains are acquired, corresponding batch sample data are acquired from each sample domain, the batch sample data corresponding to the plurality of sample domains are randomly selected from all batch sample data to serve as first sample data, the rest batch sample data corresponding to the plurality of sample domains are served as second sample data, then each first sample data is input into a target model, a feature extraction network and a discrimination network are respectively used for carrying out forward computation on each first sample data to obtain a plurality of first loss function values, backward computation is carried out to obtain a plurality of discrimination network parameters, each second sample data is then input into a target model loaded with each discrimination network parameter, forward computation is carried out on each second sample data to obtain a plurality of second loss function values, finally, the total loss function value is calculated according to each first loss function value and each second loss function value, backward computation is carried out according to the total loss function value, the operation of acquiring the plurality of first sample data and the second sample data from the plurality of sample domains is carried out in a return mode until the technical condition of finishing training is met, the machine learning process can be realized, the machine learning model can be effectively improved, the machine learning cost can be saved, and the machine learning model can be effectively developed in different models.
For better description of the technical solution provided by the embodiment of the present invention, fig. 2b is a schematic diagram of an implementation manner of a training method of a machine learning model in this embodiment, and the embodiment of the present invention may refer to the following implementation manner:
assuming that the target model is a living body detection model, firstly dividing a plurality of sample data into living body samples and non-living body samples according to label attributes; then dividing the sample data into four sample sets according to the data source attribute of each sample data, such as sample set A, sample set B, sample set C and sample set D in fig. 2B, and then adopting a data enhancement method to adjust the proportion of different types of sample data in each sample set and the proportion of sample data among each sample set so as to realize that the living sample and the non-living sample in each sample set are kept at 1:1, the number of samples between each sample set remains 1:1, and taking each sample set after adjustment as a sample domain, such as a sample domain a, a sample domain B, a sample domain C and a sample domain D in fig. 2B.
The living body detection model in the example is divided into two parts, namely a feature extraction network (Clue Extraction Network, clueExtr) and a living body discrimination network FasCls, and in one iteration training, the specific implementation steps are as follows:
Step 1: and acquiring sample data which participate in small batch training in each sample domain according to the configuration of super parameters from a data loader corresponding to different sample domains, namely batch sample data D1, D2, D3 and D4, randomly scrambling the batch sample data by using a random sequencing algorithm, selecting batch sample data which are ranked in the first two bits to participate in learning in an exploration phase, marking the batch sample data as A1 and A2 (namely first sample data), and selecting batch sample data which are ranked in the last two bits to participate in learning in a correction phase, marking the batch sample data as B1 and B2 (namely second sample data).
Step 2: in the exploration-learning phase, the following is performed for each of the first sample data A1, A2 participating in this phase: respectively carrying out forward computation on A1 and A2 through a feature extraction network, and outputting respective corresponding features Feat_A1 and Feat_A2; then, according to the corresponding characteristics of A1 and A2, respectively performing forward calculation on A1 and A2 through a living body judging network, and outputting respective classification information fascinfo_A1 and fascinfo_A2; because the living body detection task is a classification model, according to the classification information of each of A1 and A2, a cross entropy algorithm is used for calculating first Loss function values Cls_los_A1 and Cls_los_A2 corresponding to each of A1 and A2; and then, respectively carrying out backward calculation on the A1 and the A2 through a discrimination network to obtain discrimination network parameters Cls_parameters_A1 and Cls_parameters_A2 corresponding to the A1 and the A2 respectively.
After the above processing, cls_loss_a1 and cls_loss_a2 are combined into cls_loss_a by tensor addition. Then, a first twin Loss function value contrast_loss_A is calculated by a twin network Loss function according to the characteristics Feat_A1 and Feat_A2 corresponding to the A1 and the A2 respectively.
Step 3: in the deskewing phase, the following operations are performed for both the second sample data B1, B2 participating in this phase: and respectively carrying out forward computation on B1 and B2 through a feature extraction network, and outputting the corresponding features Feat_B1 and Feat_B2. The living body discrimination network loads the discrimination network parameters, performs forward computation on B1 and B2, and outputs classification information corresponding to the discrimination network parameters, namely, fascinfo_B1_A1, fascinfo_B1_A2, fascinfo_B2_A1 and fascinfo_B2_A2. Here, fascinfo_b1_a1 represents classification information corresponding to B1 when the living body discrimination network loads the discrimination network parameters corresponding to A1.
And then calculating second Loss function values Cls_los_B1_A1, cls_los_B1_A2, cls_los_B2_A1 and Cls_los_B2_A2 corresponding to the B1 and B2 respectively by using a cross entropy algorithm according to the classification information corresponding to the B1 and the B2 respectively. Wherein cls_loss_b1_a1 represents a second Loss function value corresponding to B1 when the living body discrimination network loads the discrimination network parameters corresponding to A1.
After the above processing, cls_loss_b1_a1 and cls_loss_b1_a2 are combined into cls_loss_b1 by tensor addition; combining cls_loss_b2_a1 and cls_loss_b2_a2 into cls_loss_b2 by tensor addition; cls_Loss_B1 and Cls_Loss_B2 are combined into Cls_Loss_B by tensor addition. And then, calculating a second twin Loss function value contrast_loss_B through a twin network Loss function according to the characteristics Feat_B1 and Feat_B2 corresponding to the B1 and the B2 respectively.
Step 4: the Cls_Loss_ A, cls _Loss_B is combined into Cls_Loss (namely, a first function value) through tensor addition, the contrast_Loss_ A, contrast _Loss_B is combined into contrast_Loss (namely, a second function value) through tensor addition, and the total Loss function value total of the current iteration is calculated according to the super parameter w_Cls_contrast corresponding to the target model, wherein:
TotalLoss＝Cls_Loss+w_Cls_Contrast*Contrast_Loss
then, the backward calculation of the present round of iteration is performed on the total loss function value.
Step 5: repeating the steps 1-4 until the target model converges. The condition for ending training of the target model may be: the accuracy of the training result of the target model is larger than a preset accuracy threshold, or the total loss function value is smaller than the preset loss threshold.
The preset accuracy threshold may be set to 99.9999%, the preset loss threshold may be set to 0.000001, and the specific value is preset according to the actual situation, which is not limited in this embodiment.
The method provided by the embodiment of the invention can realize the automation of the training process of the machine learning model, effectively improve the robustness of the machine learning model in different scenes and save the development cost of the machine learning model.
Example III
Fig. 3 is a schematic structural diagram of a training device for a machine learning model according to a third embodiment of the present invention. The apparatus may be implemented in software and/or hardware and may be generally integrated in a computer device. As shown in fig. 3, the apparatus includes: a sample data acquisition module 310, a first sample data input module 320, a second sample data input module 330, and a total loss function value calculation module 340.
The sample data obtaining module 310 is configured to obtain a plurality of first sample data and second sample data from a plurality of sample domains, where each sample domain includes a plurality of sample data corresponding to a same scene; the first sample data input module 320 is configured to input each piece of first sample data into a target model, perform forward computation on each piece of first sample data by using a feature extraction network and a discrimination network in the target model respectively, obtain a plurality of first loss function values, and perform backward computation to obtain a plurality of discrimination network parameters; a second sample data input module 330, configured to input each second sample data into a target model loaded with each discrimination network parameter, and perform forward computation on each second sample data to obtain a plurality of second loss function values; the total loss function value calculation module 340 is configured to calculate a total loss function value according to each first loss function value and each second loss function value, perform backward calculation according to the total loss function value, and return to perform operations of obtaining a plurality of first sample data and second sample data from a plurality of sample domains until the end training condition is satisfied.
According to the embodiment of the invention, a plurality of first sample data and second sample data are acquired from a plurality of sample domains, each first sample data is input into a target model, a feature extraction network and a discrimination network in the target model are respectively used for carrying out forward calculation on each first sample data to obtain a plurality of first loss function values, and backward calculation is carried out to obtain a plurality of discrimination network parameters; inputting each second sample data into a target model loaded with each discrimination network parameter, and performing forward calculation on each second sample data to obtain a plurality of second loss function values; according to the first loss function value and the second loss function value, the total loss function value is obtained through calculation, backward calculation is carried out according to the total loss function value, the operation of obtaining a plurality of first sample data and second sample data from a plurality of sample domains is carried out in a return mode until the end training condition is met, automation of the machine learning model training process can be achieved, robustness of the machine learning model in different scenes is effectively improved, and development cost of the machine learning model is saved.
Based on the above embodiments, the sample data obtaining module 310 may include:
A batch sample data obtaining unit, configured to obtain a plurality of sample domains, and obtain corresponding batch sample data from each sample domain;
the sample data selecting unit is used for randomly selecting batch sample data corresponding to a plurality of sample domains from all batch sample data as first sample data, and taking the batch sample data corresponding to the rest sample domains as second sample data;
the sample data dividing unit is used for acquiring a plurality of sample data and dividing each sample data into a plurality of sample sets according to a preset classification rule;
the proportion adjustment unit is used for adjusting the proportion of the sample data of different types in each sample set and the proportion of the sample data among each sample set to obtain a plurality of sample domains;
and the data acquisition unit is used for acquiring batch sample data matched with the batch threshold value from each sample domain according to the preset batch threshold value through a data loader corresponding to each sample domain.
On the basis of the above embodiments, the first sample data input module 320 may include:
and the first twin loss function value calculation unit is used for acquiring the characteristics corresponding to the first sample data output by the characteristic extraction network in the target model and calculating a first twin loss function value through a twin network loss function according to the characteristics corresponding to the first sample data.
The second sample data input module 330 may include:
and a second twin loss function value calculation unit, configured to obtain features corresponding to each second sample data output by the feature extraction network in the target model, and calculate a second twin loss function value through a twin network loss function according to the features corresponding to each second sample data.
The total loss function value calculation module 340 may include:
the parameter updating unit is used for calculating a total loss function value according to each first loss function value, each second loss function value, each first twin loss function value and each second twin loss function value, performing backward calculation according to the total loss function value and updating the parameter of the target model according to a calculation result;
a first function value calculation unit configured to tense-add each of the first loss function values and the second loss function values to obtain a first function value;
a second function value calculation unit configured to tensorily add the first twin loss function value and the second twin loss function value to obtain a second function value;
and the loss function value calculation unit is used for calculating the total loss function value according to the first function value, the second function value and the super parameter corresponding to the target model.
The training device of the machine learning model can execute the training method of the machine learning model provided by any embodiment of the invention, and has the corresponding functional modules and beneficial effects of executing the training method of the machine learning model.
Example IV
Fig. 4 is a schematic structural diagram of a computer device according to a fourth embodiment of the present invention, and as shown in fig. 4, the computer device includes a processor 410, a memory 420, an input device 430 and an output device 440; the number of processors 410 in the computer device may be one or more, one processor 410 being taken as an example in fig. 4; the processor 410, memory 420, input device 430, and output device 440 in the computer device may be connected by a bus or other means, for example in fig. 4.
The memory 420 is used as a computer readable storage medium, and may be used to store a software program, a computer executable program, and modules, such as program instructions/modules corresponding to a training method of a machine learning model in an embodiment of the present invention (for example, the sample data obtaining module 310, the first sample data input module 320, the second sample data input module 330, and the total loss function value calculating module 340 in a training device of a machine learning model). The processor 410 executes various functional applications and data processing of the computer device by running software programs, instructions and modules stored in the memory 420, i.e., implements a machine learning model training method as described above. That is, the program, when executed by the processor, implements:
Acquiring a plurality of first sample data and second sample data from a plurality of sample domains, each sample domain including a plurality of sample data corresponding to the same scene;
inputting each first sample data into a target model, performing forward calculation on each first sample data by using a feature extraction network and a discrimination network in the target model respectively to obtain a plurality of first loss function values, and performing backward calculation to obtain a plurality of discrimination network parameters;
inputting each second sample data into a target model for loading each discrimination network parameter, and performing forward calculation on each second sample data to obtain a plurality of second loss function values;
and according to the first loss function value and the second loss function value, calculating to obtain a total loss function value, carrying out backward calculation according to the total loss function value, and returning to execute the operation of acquiring a plurality of first sample data and second sample data from a plurality of sample domains until the end training condition is met.
Memory 420 may include primarily a program storage area and a data storage area, wherein the program storage area may store an operating system, at least one application program required for functionality; the storage data area may store data created according to the use of the terminal, etc. In addition, memory 420 may include high-speed random access memory, and may also include non-volatile memory, such as at least one magnetic disk storage device, flash memory device, or other non-volatile solid-state storage device. In some examples, memory 420 may further include memory remotely located relative to processor 410, which may be connected to the computer device via a network. Examples of such networks include, but are not limited to, the internet, intranets, local area networks, mobile communication networks, and combinations thereof.
The input device 430 may be used to receive input numeric or character information and to generate key signal inputs related to user settings and function control of the computer device, which may include a keyboard, mouse, and the like. The output 440 may include a display device such as a display screen.
On the basis of the above embodiments, the processor 410 is configured to obtain a plurality of first sample data and second sample data from a plurality of sample domains by:
acquiring a plurality of sample domains, and acquiring corresponding batch sample data from each sample domain;
and randomly selecting the batch sample data corresponding to the plurality of sample domains from all the batch sample data as first sample data, and taking the batch sample data corresponding to the rest plurality of sample domains as second sample data.
On the basis of the above embodiments, the processor 410 is arranged to obtain a plurality of sample fields by:
acquiring a plurality of sample data, and dividing each sample data into a plurality of sample sets according to a preset classification rule;
and adjusting the proportion of the sample data of different types in each sample set and the proportion of the sample data among each sample set to obtain a plurality of sample domains.
On the basis of the above embodiments, the processor 410 is configured to obtain corresponding bulk sample data from each sample domain by:
and acquiring batch sample data matched with the batch threshold value from each sample domain according to a preset batch threshold value through a data loader corresponding to each sample domain.
The processor 410, after deriving the plurality of discriminating network parameters, further performs the following operations:
acquiring characteristics corresponding to each first sample data output by a characteristic extraction network in the target model, and calculating a first twin loss function value through a twin network loss function according to the characteristics corresponding to each first sample data;
the processor 410, after deriving the plurality of second loss function values, further performs the following:
acquiring characteristics corresponding to each second sample data output by a characteristic extraction network in the target model, and calculating a second twin loss function value through a twin network loss function according to the characteristics corresponding to each second sample data;
the processor 410 is configured to calculate a total loss function value from each of the first loss function value and each of the second loss function value, and perform a backward calculation from the total loss function value by:
And calculating to obtain a total loss function value according to the first loss function value, the second loss function value, the first twin loss function value and the second twin loss function value, performing backward calculation according to the total loss function value, and updating the parameters of the target model according to the calculation result.
The processor 410 is configured to calculate the total loss function value from each of the first loss function value, the second loss function value, the first twin loss function value, and the second twin loss function value by:
tensor adding the first loss function value and the second loss function value to obtain a first function value;
tensor adding the first twin loss function value and the second twin loss function value to obtain a second function value;
and calculating the total loss function value according to the first function value, the second function value and the super parameter corresponding to the target model.
Example five
The fifth embodiment of the present invention further provides a computer readable storage medium, on which a computer program is stored, where the program is executed by a processor to implement the method according to any embodiment of the present invention. Of course, the computer readable storage medium provided by the embodiments of the present invention may perform the related operations in the training method of the machine learning model provided by any of the embodiments of the present invention. That is, the program, when executed by the processor, implements:
Acquiring a plurality of first sample data and second sample data from a plurality of sample domains, each sample domain including a plurality of sample data corresponding to the same scene;
inputting each first sample data into a target model, performing forward calculation on each first sample data by using a feature extraction network and a discrimination network in the target model respectively to obtain a plurality of first loss function values, and performing backward calculation to obtain a plurality of discrimination network parameters;
inputting each second sample data into a target model for loading each discrimination network parameter, and performing forward calculation on each second sample data to obtain a plurality of second loss function values;
and according to the first loss function value and the second loss function value, calculating to obtain a total loss function value, carrying out backward calculation according to the total loss function value, and returning to execute the operation of acquiring a plurality of first sample data and second sample data from a plurality of sample domains until the end training condition is met.
From the above description of embodiments, it will be clear to a person skilled in the art that the present invention may be implemented by means of software and necessary general purpose hardware, but of course also by means of hardware, although in many cases the former is a preferred embodiment. Based on such understanding, the technical solution of the present invention may be embodied essentially or in a part contributing to the prior art in the form of a software product, which may be stored in a computer readable storage medium, such as a floppy disk, a Read-Only Memory (ROM), a random access Memory (Random Access Memory, RAM), a FLASH Memory (FLASH), a hard disk or an optical disk of a computer, etc., and include several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute the method according to the embodiments of the present invention.
It should be noted that, in the embodiment of the training device of the machine learning model, each unit and module included are only divided according to the functional logic, but are not limited to the above-mentioned division, so long as the corresponding functions can be implemented; in addition, the specific names of the functional units are also only for distinguishing from each other, and are not used to limit the protection scope of the present invention.
Note that the above is only a preferred embodiment of the present invention and the technical principle applied. It will be understood by those skilled in the art that the present invention is not limited to the particular embodiments described herein, but is capable of various obvious changes, rearrangements and substitutions as will now become apparent to those skilled in the art without departing from the scope of the invention. Therefore, while the invention has been described in connection with the above embodiments, the invention is not limited to the embodiments, but may be embodied in many other equivalent forms without departing from the spirit or scope of the invention, which is set forth in the following claims.
Claims (11)
1. A method of training a machine learning model, comprising:
acquiring a plurality of first sample data and second sample data from a plurality of sample domains, each sample domain including a plurality of sample data corresponding to the same scene; the first sample data and the second sample data are picture data;
Inputting each first sample data into a target model, performing forward calculation on each first sample data by using a feature extraction network and a discrimination network in the target model respectively to obtain a plurality of first loss function values, and performing backward calculation to obtain a plurality of discrimination network parameters;
acquiring characteristics corresponding to each first sample data output by a characteristic extraction network in the target model, and calculating a first twin loss function value through a twin network loss function according to the characteristics corresponding to each first sample data;
inputting each second sample data into a target model for loading each discrimination network parameter, and performing forward calculation on each second sample data to obtain a plurality of second loss function values;
acquiring characteristics corresponding to each second sample data output by a characteristic extraction network in the target model, and calculating a second twin loss function value through a twin network loss function according to the characteristics corresponding to each second sample data;
and according to the first loss function value, the second loss function value, the first twin loss function value and the second twin loss function value, calculating to obtain a total loss function value, performing backward calculation according to the total loss function value, updating the parameters of the target model according to the calculation result, and returning to execute the operation of acquiring a plurality of first sample data and second sample data from a plurality of sample domains until the end training condition is met.
2. The method of claim 1, wherein obtaining a plurality of first sample data and second sample data from a plurality of sample domains, comprises:
acquiring a plurality of sample domains, and acquiring corresponding batch sample data from each sample domain;
and randomly selecting the batch sample data corresponding to the plurality of sample domains from all the batch sample data as first sample data, and taking the batch sample data corresponding to the rest plurality of sample domains as second sample data.
3. The method of claim 2, wherein obtaining a plurality of sample fields comprises:
acquiring a plurality of sample data, and dividing each sample data into a plurality of sample sets according to a preset classification rule;
and adjusting the proportion of the sample data of different types in each sample set and the proportion of the sample data among each sample set to obtain a plurality of sample domains.
4. The method of claim 2, wherein obtaining corresponding batches of sample data from each sample domain comprises:
and acquiring batch sample data matched with the batch threshold value from each sample domain according to a preset batch threshold value through a data loader corresponding to each sample domain.
5. The method of claim 1, wherein calculating the total loss function value from each of the first loss function value, the second loss function value, the first twin loss function value, and the second twin loss function value comprises:
tensor adding the first loss function value and the second loss function value to obtain a first function value;
tensor adding the first twin loss function value and the second twin loss function value to obtain a second function value;
and calculating the total loss function value according to the first function value, the second function value and the super parameter corresponding to the target model.
6. A computer device comprising a processor and a memory for storing instructions that, when executed, cause the processor to:
acquiring a plurality of first sample data and second sample data from a plurality of sample domains, each sample domain including a plurality of sample data corresponding to the same scene; the first sample data and the second sample data are picture data;
inputting each first sample data into a target model, performing forward calculation on each first sample data by using a feature extraction network and a discrimination network in the target model respectively to obtain a plurality of first loss function values, and performing backward calculation to obtain a plurality of discrimination network parameters;
Acquiring characteristics corresponding to each first sample data output by a characteristic extraction network in the target model, and calculating a first twin loss function value through a twin network loss function according to the characteristics corresponding to each first sample data;
inputting each second sample data into a target model for loading each discrimination network parameter, and performing forward calculation on each second sample data to obtain a plurality of second loss function values;
acquiring characteristics corresponding to each second sample data output by a characteristic extraction network in the target model, and calculating a second twin loss function value through a twin network loss function according to the characteristics corresponding to each second sample data;
and according to the first loss function value, the second loss function value, the first twin loss function value and the second twin loss function value, calculating to obtain a total loss function value, performing backward calculation according to the total loss function value, updating the parameters of the target model according to the calculation result, and returning to execute the operation of acquiring a plurality of first sample data and second sample data from a plurality of sample domains until the end training condition is met.
7. The computer device of claim 6, wherein the processor is configured to obtain the plurality of first sample data and the second sample data from the plurality of sample fields by:
Acquiring a plurality of sample domains, and acquiring corresponding batch sample data from each sample domain;
and randomly selecting the batch sample data corresponding to the plurality of sample domains from all the batch sample data as first sample data, and taking the batch sample data corresponding to the rest plurality of sample domains as second sample data.
8. The computer device of claim 7, wherein the processor is configured to obtain the plurality of sample fields by:
acquiring a plurality of sample data, and dividing each sample data into a plurality of sample sets according to a preset classification rule;
and adjusting the proportion of the sample data of different types in each sample set and the proportion of the sample data among each sample set to obtain a plurality of sample domains.
9. The apparatus of claim 7, wherein the processor is configured to obtain the corresponding batch of sample data from each sample domain by:
and acquiring batch sample data matched with the batch threshold value from each sample domain according to a preset batch threshold value through a data loader corresponding to each sample domain.
10. The apparatus of claim 6, wherein the processor is configured to calculate the total loss function value from each of the first loss function value, the second loss function value, the first twin loss function value, and the second twin loss function value by:
Tensor adding the first loss function value and the second loss function value to obtain a first function value;
tensor adding the first twin loss function value and the second twin loss function value to obtain a second function value;
and calculating the total loss function value according to the first function value, the second function value and the super parameter corresponding to the target model.
11. A computer readable storage medium having stored thereon a computer program, which when executed by a processor implements a method of training a machine learning model according to any of claims 1-5.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011591857.7A CN112633517B (en) | 2020-12-29 | 2020-12-29 | Training method of machine learning model, computer equipment and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011591857.7A CN112633517B (en) | 2020-12-29 | 2020-12-29 | Training method of machine learning model, computer equipment and storage medium |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN112633517A CN112633517A (en) | 2021-04-09 |
| CN112633517B true CN112633517B (en) | 2024-02-02 |
Family
ID=75286086
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202011591857.7A Active CN112633517B (en) | 2020-12-29 | 2020-12-29 | Training method of machine learning model, computer equipment and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112633517B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114283307B (en) * | 2021-12-24 | 2023-10-27 | 中国科学技术大学 | Network training method based on resampling strategy |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110210503A (en) * | 2019-06-14 | 2019-09-06 | 厦门历思科技服务有限公司 | A kind of seal recognition methods and device and equipment |
| CN111340850A (en) * | 2020-03-20 | 2020-06-26 | 军事科学院***工程研究院***总体研究所 | Ground target tracking method of unmanned aerial vehicle based on twin network and central logic loss |
| CN111462173A (en) * | 2020-02-28 | 2020-07-28 | 大连理工大学人工智能大连研究院 | Visual tracking method based on twin network discriminant feature learning |
| CN111612281A (en) * | 2020-06-23 | 2020-09-01 | 中国人民解放军国防科技大学 | Method and device for predicting pedestrian flow peak value of subway station and computer equipment |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| EP3731154A1 (en) * | 2019-04-26 | 2020-10-28 | Naver Corporation | Training a convolutional neural network for image retrieval with a listwise ranking loss function |
-
2020
- 2020-12-29 CN CN202011591857.7A patent/CN112633517B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110210503A (en) * | 2019-06-14 | 2019-09-06 | 厦门历思科技服务有限公司 | A kind of seal recognition methods and device and equipment |
| CN111462173A (en) * | 2020-02-28 | 2020-07-28 | 大连理工大学人工智能大连研究院 | Visual tracking method based on twin network discriminant feature learning |
| CN111340850A (en) * | 2020-03-20 | 2020-06-26 | 军事科学院***工程研究院***总体研究所 | Ground target tracking method of unmanned aerial vehicle based on twin network and central logic loss |
| CN111612281A (en) * | 2020-06-23 | 2020-09-01 | 中国人民解放军国防科技大学 | Method and device for predicting pedestrian flow peak value of subway station and computer equipment |
Also Published As
| Publication number | Publication date |
|---|---|
| CN112633517A (en) | 2021-04-09 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US20180268296A1 (en) | Machine learning-based network model building method and apparatus | |
| CN110232373A (en) | Face cluster method, apparatus, equipment and storage medium | |
| TW201939400A (en) | Method and device for determining group of target users | |
| Tang et al. | Exploring fusion strategies for accurate RGBT visual object tracking | |
| Liu et al. | Learning with nested scene modeling and cooperative architecture search for low-light vision | |
| JP6908302B2 (en) | Learning device, identification device and program | |
| CN110633786A (en) | Techniques for determining artificial neural network topology | |
| CN111309927B (en) | Personalized learning path recommendation method and system based on knowledge graph mining | |
| Chen et al. | Multi-granular mining for boundary regions in three-way decision theory | |
| CN112418320B (en) | Enterprise association relation identification method, device and storage medium | |
| WO2023103864A1 (en) | Node model updating method for resisting bias transfer in federated learning | |
| CN112906865B (en) | Neural network architecture searching method and device, electronic equipment and storage medium | |
| CN116912629B (en) | General image text description generation method and related device based on multi-task learning | |
| CN112508190A (en) | Method, device and equipment for processing structured sparse parameters and storage medium | |
| Feng et al. | Fsrf: an improved random forest for classification | |
| Senge et al. | Fast fuzzy pattern tree learning for classification | |
| CN112633517B (en) | Training method of machine learning model, computer equipment and storage medium | |
| Pajot et al. | Unsupervised adversarial image inpainting | |
| Mills et al. | Exploring neural architecture search space via deep deterministic sampling | |
| CN111667018A (en) | Object clustering method and device, computer readable medium and electronic equipment | |
| Zhang et al. | A broad generative network for two-stage image outpainting | |
| Zhuang et al. | Time-to-space progressive network using overlap skeleton contexts for action recognition | |
| CN113052208B (en) | Vision-based coal rock identification method, storage medium and electronic equipment | |
| CN115878891A (en) | Live content generation method, device, equipment and computer storage medium | |
| US11676050B2 (en) | Systems and methods for neighbor frequency aggregation of parametric probability distributions with decision trees using leaf nodes |
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 |