1. A task co-processing method for model training, which is used for AI model training, is characterized by comprising the following steps:

receiving first data, wherein the first data are sample data used for AI training, and adding the first data into a first data set;

generating second data, the second data comprising a subset of fourth data, the fourth data being a superset of an intersection of at least two first data in the first data sets, the second data further comprising one of the at least two first data forming the intersection;

generating third data, wherein the third data is a return value of the second data, and is used for evaluating the effect of a model obtained by training the second data and adding the third data into a third data set;

and generating fifth data, wherein the fifth data are model parameters, the fifth data are generated by processing the third data, and the convergence characteristics of the fifth data are evaluated.

2. The method of claim 1, wherein the generating of the third data further comprises:

constructing a joint neural network, the construction process of the joint neural network including a determination of the subset of the fourth data.

3. The method of claim 2, wherein the generating of the third data further comprises:

training the joint neural network to obtain a joint model;

one of the third data is obtained using a joint model.

4. The method according to claim 1, wherein the generating of the fifth data further comprises:

determining the fifth data from the third data set.

5. The method according to claim 5, wherein the generating of the fifth data further comprises:

and comparing the fifth data with sixth data, wherein the sixth data is a threshold value of the convergence effect of the fifth data.

6. The method of claim 1, further comprising:

and determining the convergence characteristic of the fifth data, training a joint neural network under the condition of determining convergence, and generating an optimal model set corresponding to the first data set.

7. A computer program, characterized in that it comprises means for performing the method according to any one of claims 1 to 7.

8. A computer-readable storage medium, characterized in that the computer storage medium stores program instructions that, when executed by a processor, cause the processor to perform the method of any of claims 1-7.

9. The device for executing the computer program is characterized by comprising a processing component, a storage component and a communication module component, wherein the processing component, the storage component and the communication module component are connected with each other, the storage component is used for storing data processing codes, and the communication module is used for carrying out information interaction with external equipment; the processing component is configured for invoking program code for performing the method according to any one of claims 1-7.

Background

It is currently widely recognized that Artificial Intelligence (AI-Artificial Intelligence) will be one of the most influential technologies in the twenty-first century and beyond. For AI, the core functions are embodied as an AI model, which is derived by training the samples using some AI algorithm. Therefore, the quality of one sample data tends to have a significant impact on the utility and quality of the model.

Reinforcement learning, which evolves from machine learning, uses data reinforcement to improve the quality of the sample. One typical way of enhancing data is to input more a priori knowledge, and the other typical way of enhancing data is to perform self-circulation type breadth combination or depth superposition on the data. In both modes, the sample data is hoped to be dug as much as possible, and the value of the sample data is improved.

However, for any particular sample data, the meaning and pattern included therein is limited, and thus the data enhancement method has a limited effect.

Disclosure of Invention

Therefore, the present application proposes a method, system and apparatus for solving the above problems, utilizing joint data (set) automatic notification to improve sample quality, and further improve the quality of the model obtained by training. The methods are applied to unspecified tools, equipment and systems, even a data center or a cloud service center, so that a task cooperative processing system for model training is formed. Therefore, the invention comprises the following steps:

on one hand, a task cooperative processing method for model training is provided, which comprises the following steps:

receiving first data, wherein the first data are sample data used for AI training, the first data have a corresponding AI algorithm, and the first data are added into a first data set; generating second data, the second data comprising a subset of fourth data, the fourth data being a superset of an intersection of at least two first data in the first data sets, the second data further comprising one of the at least two first data forming the intersection; generating third data, wherein the third data is a return value of the second data, the third data is used for evaluating the effect of a model obtained by using the second data as a training sample, and the third data is added into a third data set; and generating fifth data, wherein the fifth data are model parameters, the fifth data are generated by processing the third data, and the convergence characteristics of the fifth data are evaluated. Further, in the generation process of the third data, a joint neural network is constructed according to the generation mode of the subset of the fourth data; further, training the joint neural network to obtain a joint model, and obtaining a third data by using a verification set; further, in the generation process of the fifth data, the fifth data is determined according to the content of the third data set; further, according to a certain convergence threshold value, determining the convergence effect of the fifth data; further, if the fifth data does not reach the convergence effect, the fifth data is used as a model parameter, and the construction iteration of the joint neural network is carried out again; further, if the fifth data achieves the convergence effect, a joint neural network is trained, and an optimal model set of the first data set is generated.

In an environment with a server cluster or a cloud data center network, convenient and easy-to-use AI capacity from a data center is provided for a user through AI platformization and engine modeling, and a processing method with joint model task collaborative training can provide a more attractive, highly customizable and high-quality collaborative AI model for the user. The task collaborative AI modeling method comprises the following processes: decomposing the m data sets into training data and verification data respectively, initializing model parameters, and performing convergence verification on n rounds of model parameters on a data set formed by the m training data sets. For any round of convergence verification operation, the process is as follows: for example, in the k-th round, k training data sets are used for constructing a joint neural network set and generating a joint model set, wherein k is greater than 0 and not less than m; generating intersections among data for the first k data according to a data disassembling and combining mode determined by the model parameters, and generating a superset of each intersection; performing disassembly and combination of metadata item granularity on each superset, and constructing a combined neural network according to a disassembly and combination mode to obtain a combined neural network set; training each joint neural network by using a training data set to obtain a corresponding joint model so as to obtain a joint model set; and verifying each combined model by using the verification data set to obtain a corresponding return value, thereby obtaining a return value set. In practice, the number of the joint neural networks is determined by the number of the inter-data joint modes, and thus the number of the joint model is determined. Namely: and obtaining a combined neural network through the k-th round of data disassembly and combination, obtaining a k-th round of combined model through the k-th round of combined neural network, and obtaining a k-group set of return values through the total k-round combined model. And finally, calculating to obtain new model parameters by using a set of the total m groups of return value sets, and comparing convergence results of the model parameters of the m rounds and the previous m rounds to determine whether the combined model reaches the optimum. Under the condition that the model parameters are determined to not reach the convergence effect, continuously and iteratively constructing a combined neural network and verification combined AI model; and under the condition that the model parameters are determined to be optimal, distributing a collaborative learning task required by the combined model for collaborative learning and model training.

Therefore, the product and service system comprising part or all of the methods and steps is implemented, a higher-quality AI model can be provided by performing model training based on a model task capable of being trained cooperatively, and a more flexible and highly customized model output result is provided for the same AI algorithm, so that the AI has the capability of boosting more convenient cloud application and big data application, and the popularization of the cloud application and the big data application is accelerated.

On the other hand, a task cooperative processing device for model training is provided, the device comprising:

data-1: certain data for training an AI algorithm under a joint modeling and task cooperation method is shown;

a plurality of data-2: and a plurality of second data used for training a plurality of AI algorithms under the joint modeling and task cooperation method are illustrated. The first data and the plurality of second data may be stored in a data storage device, a memory module, or a memory system providing an external access interface. It should be noted that the first and second components described in 101 and 102 are not strictly distinguished, and one or more described in 101 and 102 do not represent a specific difference, but merely for convenience of description;

a model training unit: the unit is used for algorithm training to output a corresponding model;

a joint modeling unit: the unit is used for joint modeling and collaborative training to finally output an optimal joint model;

the data center station: the middle station completes various conversion and processing operations of data to cooperate and complete cooperative and joint enhancement of the data. Specifically, the middle station includes a corresponding data access interface, an acquisition unit, a disassembly unit, a presentation unit, and a combination unit, which respectively provide the operation processes of collection, disassembly, Embedding presentation, combination, and the like of the data to be cooperatively trained.

The interface and the module provided by the invention together with other units, modules, related platforms and related engines required by the actual implementation of a product complete model training for joint modeling and collaborative training based on a data set formed by a plurality of sample data (or training data or verification data or both training data and verification data) data, thereby realizing a task collaborative device for model training. The expression is as follows: the task cooperation device decomposes the m data sets into training data and verification data respectively, initializes model parameters, and conducts convergence verification of n-round model parameters on the data set formed by the m training data sets. For any round of convergence verification operation, the process is as follows: for example, in the k-th turn, the task cooperation device uses k training data sets to construct a joint neural network set and generate a joint model set, where k is greater than 0 and not less than m; according to a data disassembly and combination mode determined by the model parameters, the task cooperation device generates an intersection among data of the first k data and generates a superset of the intersection; the task cooperation device conducts disassembly and combination of the metadata item granularity on the superset, and constructs a combined neural network set according to the disassembly and combination mode, so that the combined neural network set is obtained; the task cooperation device trains each joint neural network by using the training data set to obtain a corresponding joint model, so that a joint model set is obtained; and the task cooperation device verifies each combined model by using the verification data set to obtain a corresponding return value, so that a return value set is obtained. In practice, the number of the joint neural networks is determined by the number of the inter-data joint modes, and thus the number of the joint model is determined. Namely: and obtaining a combined neural network through the k-th round of data disassembly and combination, obtaining a k-th round of combined model through the k-th round of combined neural network, and obtaining a k-group set of return values through the total k-round combined model. And finally, the task cooperative device calculates a new model parameter by using a set of the total m groups of return value sets, and compares the convergence results of the model parameters of the m rounds and the previous m rounds to determine whether the combined model reaches the optimum. Under the condition that the model parameters are determined not to reach the convergence effect, the task cooperation device continues to iterate to construct a joint neural network and verify a joint AI model; and under the condition that the model parameters are determined to be optimal, the task cooperation device allocates the cooperation learning tasks required by the joint model to carry out cooperation learning and model training. .

Therefore, the product and service system with the functional device can provide a higher-quality AI model by performing model training based on the model task capable of being trained in coordination, and provides a highly customized and more flexible model output result for the same AI algorithm, so that the AI has the capability of boosting more convenient cloud application and big data application, and the popularization of the cloud application and the big data application is accelerated.

In another aspect, a computer-readable storage medium is provided, which stores program instructions that, when executed by a processor, the processor (respectively) has implementation procedures to perform the above-described method.

In another aspect, an apparatus for management is provided that includes a storage component, a processing component, and a communication component, the storage component, the processing component, and the communication component being interconnected. The storage component is used for storing data processing codes, and the communication component is used for carrying out information interaction with external equipment; the processing component is configured to invoke program code, each to perform the functions described above with respect to the apparatus.

Drawings

In order to more clearly illustrate the technical solution of the present invention and to more clearly illustrate the elements, modes and processes for achieving the objects of the present invention, the following drawings are provided for illustrating the embodiments of the present invention:

FIG. 1 is a system diagram illustrating the cooperative processing of model training tasks according to the present invention;

FIG. 2 is a system diagram illustrating the cooperative processing of model training tasks according to the present invention;

FIG. 3 is a system diagram illustrating the cooperative processing of model training tasks according to the present invention;

FIG. 4 is one of the schematic diagrams of the product data cooperatively processed by the model training task proposed by the present invention;

FIG. 5 is one of the schematic diagrams of the product data cooperatively processed by the model training task proposed by the present invention;

FIG. 6 is one of the schematic diagrams of the product data cooperatively processed by the model training task proposed by the present invention;

FIG. 7 is one of the operation execution flows of the model training task co-processing proposed by the present invention;

FIG. 8 is one of the operation execution flows of the model training task co-processing proposed by the present invention;

FIG. 9 is one of the operation execution flows of the model training task co-processing proposed by the present invention.

Detailed Description

The embodiments of the present invention will be described below with reference to the drawings.

The terms "first", "second", and "third", etc. in the description and claims of this application and in the accompanying drawings are used for distinguishing between different objects and not for describing a particular order. Furthermore, "include" and "have," as well as any variations thereof, are intended to cover non-exclusive inclusions. For example, a process, method, system, article, or apparatus that comprises a list of steps or elements is not limited to only those steps or elements listed, but may alternatively include other steps or elements not listed, or inherent to such process, method, article, or apparatus.

Reference herein to "an embodiment" means that a particular feature, structure, or characteristic described in connection with the embodiment can be included in at least one embodiment of the application. The appearances of the phrase in various places in the specification are not necessarily all referring to the same embodiment, nor are separate or alternative embodiments mutually exclusive of other embodiments. It is explicitly and implicitly understood by one skilled in the art that the embodiments described herein can be combined with other embodiments.

As used in this application, the terms "server," "device," "apparatus," "unit," "component," "module," "system," and the like are intended to refer to a computer-related entity, either hardware, firmware, a combination of hardware and software, or software in execution. For example, a server may be, but is not limited to, a processor, a data processing platform, a computing device, a computer, two or more computers, or the like; a unit may be, but is not limited to being, a process running on a processor, a runnable object, an executable, a thread of execution, or any other executable computer program. One or more units may reside within a process and/or thread of execution and a unit may be localized on one computer and/or distributed between 2 or more computers. In addition, these units may execute from various computer readable media having various data structures stored thereon. The elements may communicate by way of local and/or remote processes based on a signal having one or more data packets (e.g., data from two elements interacting with another element in a local system, distributed system, and/or across a network, such as the internet with other systems by way of the signal).

First, some terms in the present application are explained so as to be easily understood by those skilled in the art. The terms listed include the following:

(1) cloud computing: namely Cloud Computing, refers to a new Computing paradigm that has the advantages of integration, connectivity in a network environment, and the ability to provide Computing, storage, and even software to users in a service fashion. The difference between the new computing paradigm and the old computing paradigm is that, for the user, the new computing paradigm has no visible fixed form or even no resource-free state, so the new computing paradigm is called cloud computing;

(2) artificial intelligence: the intelligent simulation method is characterized in that the method is an Artificial Intelligence, AI for short, and is a general name of a method, technology, software, hardware and a system for simulating human Intelligence through a computing system;

(3) machine learning: machine learning is an important branching technique in the field of AI. Machine learning extracts data patterns from the sample data in order to make the best possible predictions of the application data. From the current development results, machine learning is divided into supervised learning, unsupervised learning and reinforcement learning;

(4) algorithm-sample-model-return value: this is three important concepts of machine learning. The algorithm is a priori guidance, and different machine learning types determine the amount of priori knowledge of the algorithm; the priori knowledge needs a certain amount of data to convert and verify the obtained prediction capability, and the certain amount of data is called as samples or sample data; the algorithm finds some ability to predict and process future data in the value space provided by the sample data, and the machine representation of this ability is the model. In general, a sample is divided into a training sample and a verification sample; and the model is verified by a verification sample, and the obtained result is a return value. The return value can be called as the return value of the sample or the return value of the model;

(5) a joint neural network: the neural network is formed by aggregating a plurality of neural networks with a common parameter module, all inputs of all the neural networks are used as inputs, all outputs of all the neural networks are used as outputs, multi-task simultaneous learning is realized through parameter sharing of the common module, semantic representation (Embedding) of data is more accurate and covers more information, and therefore the training effect of each neural network is improved.

Next, the objective problem of the present invention and a technical method for solving the objective problem are summarized. With the development of AI applications, people have raised demands on AI in terms of high quality, ease of use, and convenience. The traditional method for obtaining the AI model by training based on the specific sample limits the flexibility of AI. Under the realistic condition that sample data is bound to be limited and an AI algorithm is scarce, in order to solve the contradiction, the invention provides a method for improving the flexibility of the AI model and the quality of the AI model by an automatic intelligent model training task cooperation method, thereby improving the usability and convenience of AI application and facilitating the popularization and promotion of the AI in a larger range.

The invention will be further explained with reference to the drawings. Wherein:

fig. 1 is one of the system components of the present invention. The figure illustrates a compositional relationship with respect to joint modeling and task orchestration functionality implementation. Wherein:

101-data-1: certain data for training an AI algorithm under a joint modeling and task cooperation method is shown;

102-multiple data-2: and a plurality of second data used for AI algorithm training under the joint modeling and task cooperation method are shown. The first data and the plurality of second data may be stored in a data storage device, a memory module, or a memory system providing an external access interface. It should be noted that the first and second components described in 101 and 102 are not strictly distinguished, and one or more described in 101 and 102 do not represent a specific difference, but merely for convenience of description;

103-model training unit: the unit is used for algorithm training to output a corresponding model;

104-joint modeling unit: the unit is used for joint modeling and task cooperation so as to finally output an optimal joint model;

105-data station: the middle station completes various conversion and processing operations of data to cooperate and complete coordination and enhancement of the data. Specifically, the middle station includes a corresponding data access interface, an acquisition unit, a disassembly unit, a presentation unit, and a combination unit, which respectively provide the operation processes of collection, disassembly, Embedding presentation, combination, and the like of the data to be cooperatively trained.

Fig. 2 is one of the system components of the present invention. The diagram illustrates a compositional relationship with respect to joint modeling and task collaborative implementation. Wherein:

201-data-1: certain data for training an AI algorithm under a joint modeling and task cooperation method is shown;

202-multiple data-2: and a plurality of second data used for AI algorithm training under the joint modeling and task cooperation method are shown. The first data and the plurality of second data may be stored in a data storage device, a memory module, or a memory system providing an external access interface. It should be noted that the first and second descriptions 201 and 202 are not strictly distinguished, and one or more descriptions 201 and 202 do not represent specific differences, but merely for convenience of description;

203-model training unit: the unit is used for algorithm training to output a corresponding model;

204-joint modeling unit: the unit is used for joint modeling and task cooperation so as to finally output an optimal joint model;

205-data storage interface: the interface is used for completing the access operation of the required data;

206-data acquisition unit: the unit is used for operations such as original acquisition of data maintained and managed by a data center station;

207-data disassembly unit: the unit is used for disassembling data needed by data combination;

208-data representation unit: the unit is used for representing the required operation realization for data Embedding appearing in the data combination process;

209-data join unit: the unit is used for data association operations required for data association.

Fig. 3 is one of the system components of the present invention. The diagram illustrates further partitioning of joint modeling in conjunction with tasks to achieve desired functional composition. Wherein:

311-data interaction unit: the unit is used for providing data interaction and data control between the application layer and the middle station;

312-application service unit: the unit is used for providing service response to data operation requirements of other aspects in the application;

313-application acquisition unit: the unit is used for providing acquisition enabling and data acquisition for the application;

321-federated policy Unit: the unit is used for providing a combination strategy required by combination with management data;

322-model training unit: the unit is used for executing the training process of the algorithm model;

323-joint modeling unit: the unit is used for performing a reinforcement learning process required by the training process;

324-model evaluation unit: the unit is used for evaluating the obtained combined model, and evaluating and using the verification data subset;

331-data acquisition unit: the unit is used for managing the collected data;

332-data association unit: the unit is used for carrying out relevance analysis on the disassembled data;

333-data disassembly unit: the unit is used for disassembling each data;

334-data representation unit: the unit is used for representing the required operation realization for data Embedding appearing in the data combination process;

341-AI modeling Engine: the engine is used for providing operation support required by algorithm modeling;

342-big data Engine: the engine is used to provide the capability and service support needed for other processing of data.

Fig. 4 is a schematic diagram of product data according to the present invention. This figure illustrates the internal representation of one of the data used by the product implementing the inventive method with other data in the processes of mapping, unpacking and combining. Wherein:

401-data schematic 1: certain data for training an AI algorithm under a joint modeling and task collaborative training method is shown;

402-multiple data schematic 2: a plurality of second data used for training the AI algorithm under the joint modeling and task collaborative training method are shown;

403-schematic of data 1 corresponding to metadata item: the metadata items corresponding to the data are illustrated;

404-schematic of multiple data 2 corresponding metadata items: this illustrates a metadata item for a plurality of data;

405-metadata schema for data 1: the illustrating complete metadata corresponding to one of the data;

406-metadata set schema for multiple data 2: the metadata set is formed by metadata corresponding to a plurality of data respectively, and metadata items which may be the same exist among all subsets of metadata in the metadata set;

407-authentication data for data 1: this illustrates the verification data divided by data 1;

408-training data for data 1: this illustrates training data divided by data 1;

409-a plurality of data 2 form a data set: the illustration is a subset of corresponding portions of data in the plurality of data sets;

it should be noted that: in one aspect, the one or more data layers 401 and 402 are illustrated in a quantity that is merely ambiguous, and the other metadata layer and the data layer illustrate a mapping relationship; on the other hand, no matter 403-406 or 407-409 are used for limiting the width and depth of the data; in yet another aspect, the data relationship of the figure is simply illustrative and not meant to be a specific limitation on the implementation of the invention.

Fig. 5 is a schematic diagram of product data according to the present invention. This figure illustrates (assuming the existence of) a sample product implementing the core method of the invention: the occupation and skill specialties are directly judged through the photos. Firstly, the product mainly uses data of a resume website as a sample to train to obtain an algorithm model, and then the model is used for identifying the occupation and skill specialties of a master of a target picture; the product after the core method of the invention is implemented has the following components: taking data of the resume website as one sample data, taking data available from other websites as the second sample data, obtaining information of different categories by combing the information of other websites, and under the condition that the types can correspond to each other, performing joint modeling and task collaborative training on the sample data of the local online shop and the sample data of other websites to obtain a new joint recognition model of 'picture- > occupation'. Wherein:

501-illustrating one of sample data used by resume websites implementing products to train models, namely resume data;

502-news data, forum data, academic data, etc. illustrating various sample data for training models for implementing news/forum/academic, etc. of products;

503-item illustrating metadata describing resume data;

504-items illustrating metadata describing the news/forum/academy three categories of data;

505-illustrates metadata describing resume data;

506-illustrates metadata describing the news/forum/academy three types of data;

507-illustrates validation data in the resume dataset;

508-illustrates training data in the resume dataset used for model training;

509-illustrates the joint training data from three types of news/forum/academia.

Fig. 6 is one of the schematic diagrams of product data for implementing the present invention. The graph illustrates the relationship of the joint training task by performing a disassembly and a joint on the data to be treated and deploying the joint based on the disassembly and the joint. Wherein:

601-schematically one of the data (shown as data-1) processed by a certain product implementing the technical method of the invention, this data comprising a subset of data to be joined with another data (shown as data-i);

602-illustrates a second of the data (illustrated as data-i) processed by a product implementing the inventive technical method, the data comprising a subset of data to be joined with another data (illustrated as data-1);

611-data-10 which shows the result of the disassembly of data-1 of a certain product implementing the technical method of the invention, i.e. it is explicitly identified that some part of data will be jointly trained with other data;

612-data showing the result data-11 after the disassembly of data-1 of a certain product implementing the technical method of the invention, namely performing combined training based on the data;

613-data showing the result of a product implementing the technical method-i 1 after the disassembly is completed, i.e. performing joint training based on the data;

614-data of a certain product implementing the technical method of the invention-i result data after completion of the disassembly-i 0, i.e. it is explicitly identified that a certain part of data is to be trained in combination with other data;

621-illustrating a training task a corresponding to data-10 of a certain product implementing the technical method of the invention;

622-data illustrating a certain product implementing the technical method of the invention-training task B corresponding to 11;

623-data of a certain product implementing the technical method of the invention-training task B' corresponding to i 1;

624-data of a certain product implementing the technical method of the invention-i 0.

It should be noted that: first, fig. 6 is only a schematic diagram of a part of features of a product for implementing the core method of the invention, not all features of the product, and not any limitation of the method and features of the invention; secondly, the number of data processed by the product implementing the core method of the invention may be more than two, and only two of the data are illustrated here; thirdly, the three stages of the data joint training, namely, the conversion and the corresponding relation from the data to the data disassembly and from the data disassembly to the joint training task deployment are illustrated, and other operations and corresponding relations are not illustrated in the figure.

Fig. 7 is one of the operation execution flow charts for implementing the present invention. The figure illustrates a process of combining neural network construction and model training by m cycles, wherein m represents the number of data to be combined, and k represents the k-th cycle in the m cycles. Wherein:

10A-adding the kth data, and preparing the construction of the joint neural network of the kth round: the operation is used for adding the kth data into a data set of joint training to construct a joint neural network;

10B-determine the joint objective of k: the operation is used for determining target data to be combined by newly added kth data according to the reinforcement learning model parameters;

10C-disassemble the kth data: the operation is to disassemble the kth data according to the model parameters;

10D-extracting metadata items common among the data: the operation is used for identifying a common metadata item among data based on disassembled data after the k-th data is disassembled;

10E-denote the data items of the common metadata as Embedding: the operation is used for identifying common metadata items among the disassembled data, extracting corresponding data according to the common items and forming an Embedding representation;

10F-building a joint neural network from the common representation: this operation is used to build a joint neural network on top of the aforesaid Embedding of the common data. In this way, multiple common metadata items may form multiple federated neural networks;

10G-Joint training A training model was obtained: the operation is used for deploying a joint training task by using the constructed joint neural network, and loading and training corresponding training data to obtain a joint model. In this way, a plurality of joint neural networks can obtain a plurality of joint models;

10H-obtaining the corresponding return value of the combined model: the operation is used for deploying tasks by using the joint model obtained through the training, and loading a corresponding verification data set, so that a return value set of the joint model is obtained. In this way, multiple federated models can result in a set of reward values.

Fig. 8 is one of the operation execution flow charts for implementing the present invention. The figure illustrates a process of performing a joint neural network iteration based on the set of return values and finally obtaining an optimal training model result. Wherein:

20A-initialization and start first round iteration: the operation is used for starting iteration of the joint neural network, such as necessary initialization of parameters of a reinforcement learning model, setting iteration rounds and the like;

20B-the way of disassembling and uniting the data: the operation is used for a data disassembly and combination strategy determined by model parameters and other factors to obtain a data disassembly mode;

20C, constructing a joint neural network, training a joint model, and obtaining a model return value: this operation is used to perform the k-th round of the cyclic operation process of the joint neural network-joint training model-model return value shown in fig. 7, so as to obtain a return value set;

20D-preparation of k +1 iterations: the operation is to prepare to add the (k + 1) th data to the iterative process if k is less than m;

20E-building a reinforcement learning model: the operation is used for constructing a reinforcement learning model in a joint mode of a data set;

20F-reinforcement learning model training: the operation is used for arranging a joint training task according to the determined data joint mode and the reinforcement learning model, so that a training data set is trained to obtain a joint learning model;

20G-updating and determining parameters of the reinforcement learning model: the operation is used for updating the reinforcement learning model parameters according to the verification result and judging whether to carry out the iteration of the joint learning according to the convergence condition of the model parameters;

20J-end training: this operation is to end the joint training if the aforementioned determination result is convergence.

Fig. 9 is one of the operation execution flow charts for implementing the present invention. This figure illustrates the process of obtaining a complete joint training model. Wherein:

30A-receive and extract dataset: the operation is used for obtaining a data set to be subjected to joint training, wherein the data set is composed of a plurality of data;

30B-divide each data into training data and validation data: the operation is used for dividing each data in the data to obtain respective training data and verification data;

30C-initializing reinforcement learning model parameters: this operation is used to initialize reinforcement learning model parameters;

30D-load the splitting and combining strategy of each data: the operation is used for acquiring and loading the disassembling and combining strategies of each data, and the strategies are used for guiding the data disassembling process and the data combining process required by the subsequent iteration process;

30E-iterative training: this operation is used to invoke the execution of the operational process shown in FIG. 8, resulting in an optimal learning model;

30F-output learning model: the operation is used for obtaining a learning model obtained through combined training and reinforcement learning model training;

30G-deployment learning model: this operation is used to deploy the final learning model.

In this application, the units described as separate parts may or may not be physically separate, and parts displayed as units may or may not be physical units, and may be located in a single network node, or may be distributed on multiple network nodes. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of the embodiment of the present invention.

In addition, according to specific constraints and implementation requirements, functional components in the embodiments of the present application may be integrated into one component, or each component may exist alone physically, or two or more components may be integrated into one component. The integrated components can be realized in a form of hardware or a form of software functional units.

The integrated components, if implemented in the form of software functional units and sold or used as a stand-alone product, may be stored in a computer readable storage medium. Based on such understanding, the technical solution of the present invention essentially contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product stored in a storage medium and including instructions for causing one or more computer devices (which may be personal computers, servers, or network devices) to execute all or part of the steps of the method according to the embodiments of the present invention. And the aforementioned storage medium includes: a U-disk, a removable hard disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), a magnetic disk or an optical disk, and other various media capable of storing program codes.

While the invention has been described with reference to specific embodiments, the scope of the invention is not limited thereto, and those skilled in the art can easily conceive various equivalent modifications or substitutions within the technical scope of the invention. Therefore, the protection scope of the present application shall be subject to the protection scope of the claims.

It should be understood that, in the various embodiments of the present application, the serial numbers of the above-mentioned processes do not mean a strict order of execution, and the execution order of the processes should be determined by their functions and inherent logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention. While the present application has been described herein in conjunction with various embodiments, other variations to the disclosed embodiments may be understood and effected by those skilled in the art in practicing the present application as claimed herein.