KR102473613B1 - Techniques for distributed learning of artificial intelligence models - Google Patents

Abstract

According to one embodiment of the present disclosure, a computer program stored in a computer readable storage medium is disclosed. The computer program includes instructions for causing a processor of a master node to perform the following steps: receiving a distributed resource request signal for model learning from a client; determining a node to perform learning of the model among the master node and one or more subnodes included in one cluster, upon receiving the distributed resource request signal; obtaining learning result information according to learning of the model, and updating the model based on the learning result information; and when the updated model satisfies a preset condition, completing learning of the model and transmitting the updated model to the client.

Description

Distributed learning technique of artificial intelligence model {TECHNIQUES FOR DISTRIBUTED LEARNING OF ARTIFICIAL INTELLIGENCE MODELS}

The present disclosure relates to an information processing method using a computing device, and more specifically, to a technique for providing a distributed learning technique for an artificial intelligence model.

An artificial intelligence (AI) system is a computer system that implements human-level intelligence, and unlike existing rule-based smart systems, machines learn, judge, and become smarter on their own. The more AI systems are used, the higher the recognition rate and the more accurate understanding of user preferences, so existing rule-based smart systems are gradually being replaced by deep learning-based AI systems.

Before these AI systems can be used in various fields, AI models must be trained.

Meanwhile, large resources of a computing device may be required to process learning of an artificial intelligence model. However, difficulties (eg, high cost, etc.) may occur in securing large resources of a computing device in order to process learning of an artificial intelligence model for an individual or small-scale company that wants to implement an artificial intelligence system.

Therefore, there may be a demand for a distributed learning technique of an artificial intelligence model that distributes and learns the learning of an artificial intelligence model requiring a large amount of resources.

Korean Patent Registration No. 10-1944090 discloses a learning method of an artificial intelligence model used in a specific field (task matching).

The present disclosure has been made in response to the aforementioned background art, and is intended to provide a distributed learning technique for an artificial intelligence model.

The technical problems of the present disclosure are not limited to the technical problems mentioned above, and other technical problems not mentioned will be clearly understood by those skilled in the art from the following description.

According to an embodiment of the present disclosure for solving the above problems, according to an embodiment of the present disclosure, a computer program stored in a computer readable storage medium, the computer program causes a processor of a master node to: and instructions for performing steps including: receiving a distributed resource request signal for learning a model from a client; determining a node to perform learning of the model among the master node and one or more subnodes included in one cluster, upon receiving the distributed resource request signal; obtaining learning result information according to learning of the model, and updating the model based on the learning result information; and when the updated model satisfies a preset condition, completing learning of the model and transmitting the updated model to the client.

In addition, upon receiving the distributed resource request signal, the step of determining a node to perform learning of the model among the master node and one or more subnodes included in one cluster, included in the distributed resource request signal recognizing a task for learning the model; Recognizing a resource of each of the one or more subnodes to distribute the task; Dividing the task into subtasks based on resources of each of the one or more subnodes; and determining a node to perform the sub-task.

In addition, the task may be independently performed by the master node and each of the one or more subnodes, or may be performed dependently by the master node and one or more subnodes.

In addition, upon receiving the distributed resource request signal, the step of determining a node to perform learning of the model among the master node and one or more subnodes included in one cluster, based on the distributed resource request signal determining learning setting information; and determining a node to perform learning of the model among the master node and one or more subnodes based on the learning setting information.

In addition, the steps may include: recognizing information about a job performed using the model included in the distributed resource request signal; and determining optimal learning setting information based on the information about the job.

In addition, the step of obtaining learning result information according to learning of the model and updating the model based on the learning result information includes determining optimal learning setting information based on the learning result information. can

In addition, in the step of determining optimal learning setting information based on the learning result information, the performance of the model learned in the master node and each of the one or more subnodes is determined, and each is based on the performance of each model. Determining optimal learning setting information among learning setting information matched to the model of; may include.

In addition, determining the performance of the model learned in the master node and each of the one or more subnodes, and determining optimal learning setting information among learning setting information matched to each model based on the performance of each model. obtaining a plurality of sub-learning result information according to the learning of the model performed in the master node and each of the one or more sub-nodes and sub-learning setting information corresponding to the plurality of sub-learning result information; updating the model into a plurality of sub-models using the plurality of sub-learning result information, and determining performance of each of the plurality of sub-models using a test data set; recognizing first sub-learning result information used to update a specific model having the highest performance among the plurality of sub-models; and determining first sub-learning setting information corresponding to the first sub-learning result information as the optimal learning setting information.

In addition, acquiring learning result information according to learning of the model and updating the model based on the learning result information may include a first learning result according to learning of the model performed in each of the one or more subnodes. receiving information; obtaining second learning result information according to learning of the model performed in the master node; and obtaining the learning result information by summing the first learning result information and the second learning result information, and updating the model based on the learning result information.

In addition, the step of obtaining learning result information according to learning of the model and updating the model based on the learning result information, according to the learning of the model performed in each of the master node and the one or more subnodes. obtaining a plurality of sub-learning result information; updating the model into a plurality of sub-models using the plurality of sub-learning result information, and determining performance of each of the plurality of sub-models using a test data set; and determining learning result information of a model determined based on the performance of each of the plurality of sub-models as the learning result information, and determining the model determined based on the performance as the updated model. .

In addition, when the updated model satisfies a preset condition, completing learning of the model and transmitting the updated model to the client may include obtaining evaluation information of the updated model; Based on the learning target information and the evaluation information included in the distributed resource request signal, determining whether the predetermined condition is satisfied; may include.

In addition, the step of determining whether or not the preset condition is satisfied based on the learning target information and the evaluation information included in the distributed resource request signal includes the evaluation information than the target performance of the model included in the learning target information. and recognizing that the preset condition is satisfied when the performance of the updated model included in is high.

In addition, when the updated model satisfies a preset condition, completing learning of the model and transmitting the updated model to the client includes transmitting the optimal learning setting information together with the updated model. The step of doing; may include.

In addition, the distributed resource request signal for learning the model uses identification information capable of identifying the client, learning setting information required for learning the model, learning target information that is a learning goal of the model, and the model to be learned. It may include at least one of information about a job performed by doing so.

In addition, the learning setting information may include hyperparameter information related to the model.

In addition, the learning result information may include at least one of version information of the learned model, weights of the learned models, weight changes according to the learning, and information about a learning method of the model.

In addition, the weight of the learned model may reflect knowledge extracted from a training data set used to train the model in the master node and the one or more subnodes.

The technical solutions obtainable in the present disclosure are not limited to the above-mentioned solutions, and other solutions not mentioned will become clear to those skilled in the art from the description below. You will be able to understand.

According to some embodiments of the present disclosure, it is possible to provide a distributed learning technique of an artificial intelligence model capable of performing learning of an artificial intelligence model without being limited to the size of resources of a computing device.

Effects obtainable in the present disclosure are not limited to the effects mentioned above, and other effects not mentioned will be clearly understood by those skilled in the art from the description below. .

Various aspects are now described with reference to the drawings, wherein like reference numbers are used to collectively refer to like elements. In the following embodiments, for explanation purposes, numerous specific details are set forth in order to provide a thorough understanding of one or more aspects. However, it will be apparent that such aspect(s) may be practiced without these specific details.
1 is a schematic diagram of a distributed learning system for an artificial intelligence model according to some embodiments of the present disclosure.
2 is a flowchart illustrating an example of a method for a master node to perform distributed learning of an artificial intelligence model according to some embodiments of the present disclosure.
3 is a flowchart illustrating an example of a method in which a master node divides tasks for distributed learning of an artificial intelligence model according to some embodiments of the present disclosure.
4 is a flowchart illustrating an example of a method for a master node to update an artificial intelligence model based on learning result information according to some embodiments of the present disclosure.
5 is an exemplary diagram briefly illustrating a network structure of artificial intelligence models that can be learned according to some embodiments of the present disclosure.
6 shows a general schematic diagram of an example computing environment in which embodiments of the present disclosure may be implemented.

Various embodiments are now described with reference to the drawings. In this specification, various descriptions are presented to provide an understanding of the present disclosure. However, it is apparent that these embodiments may be practiced without these specific details.

The terms “component,” “module,” “system,” and the like, as used herein, refer to a computer-related entity, hardware, firmware, software, a combination of software and hardware, or an execution of software. For example, a component may be, but is not limited to, a procedure, processor, object, thread of execution, program, and/or computer running on a processor. For example, both the server and the application running on the server can be components. One or more components may reside within a processor and/or thread of execution. A component can be localized within a single computer. A component may be distributed between two or more computers. Also, these components can execute from various computer readable media having various data structures stored thereon. Components may be connected, for example, via signals with one or more packets of data (e.g., data and/or signals from one component interacting with another component in a local system, distributed system) to other systems and over a network such as the Internet. data being transmitted) may communicate via local and/or remote processes.

In addition, the term “or” is intended to mean an inclusive “or” rather than an exclusive “or”. That is, unless otherwise specified or clear from the context, “X employs A or B” is intended to mean one of the natural inclusive substitutions. That is, X uses A; X uses B; Or, if X uses both A and B, "X uses either A or B" may apply to either of these cases. Also, the term "and/or" as used herein should be understood to refer to and include all possible combinations of one or more of the listed related items.

Also, the terms "comprises" and/or "comprising" should be understood to mean that the features and/or components are present. However, it should be understood that the terms "comprises" and/or "comprising" do not exclude the presence or addition of one or more other features, elements, and/or groups thereof. Also, unless otherwise specified or where the context clearly indicates that a singular form is indicated, the singular in this specification and claims should generally be construed to mean "one or more".

Also, the terms “information” and “data” used herein may often be used interchangeably with each other.

In addition, the term “at least one of A or B” should be interpreted as meaning “when only A is included”, “when only B is included”, and “when A and B are combined”.

Those skilled in the art will further understand that the various illustrative logical blocks, components, modules, circuits, means, logics, and algorithm steps described in connection with the embodiments disclosed herein may be implemented using electronic hardware, computer software, or combinations of both. It should be recognized that it can be implemented as To clearly illustrate the interchangeability of hardware and software, various illustrative components, blocks, configurations, means, logics, modules, circuits, and steps have been described above generally in terms of their functionality. Whether such functionality is implemented in hardware or as software depends on the particular application and design constraints imposed on the overall system. Skilled artisans may implement the described functionality in varying ways for each particular application. However, such implementation decisions should not be interpreted as causing a departure from the scope of this disclosure.

The suffixes "module" and "unit" for the components used in the following description are given or used interchangeably in consideration of ease of writing the specification, and do not have meanings or roles that are distinct from each other by themselves.

Objects and effects of the present disclosure, and technical configurations for achieving them will become clear with reference to embodiments described later in detail in conjunction with the accompanying drawings. In describing the present disclosure, if it is determined that a detailed description of a known function or configuration may unnecessarily obscure the gist of the present disclosure, the detailed description will be omitted. In addition, terms to be described later are terms defined in consideration of functions in the present disclosure, which may vary according to intentions or customs of users or operators.

However, the present disclosure is not limited to the embodiments disclosed below and may be implemented in a variety of different forms. These embodiments are provided only to make this disclosure complete and to completely inform those skilled in the art of the scope of the disclosure, and the disclosure is only defined by the scope of the claims. . Therefore, the definition should be made based on the contents throughout this specification.

The scope of rights to the method in the claims of the present disclosure is caused by the functions and features described in each step, and unless the order of precedence is specified in each step constituting the method, the scope of the claims It is not affected by the order of description of each step in . For example, in a claim described in a method including steps A and B, even if step A is described before step B, the scope of rights is not limited to the fact that step A must precede step B.

1 is a schematic diagram of a distributed learning system for an artificial intelligence model according to some embodiments of the present disclosure. The terms "artificial intelligence model" and "model" used in the following description may often be used interchangeably with each other.

Referring to FIG. 1 , a cluster 1000 may include a master node 100 and one or more sub nodes 200 . Here, the cluster 1000 may refer to a set of nodes for dispersively performing a specific task (eg, learning of an artificial intelligence model).

Specifically, a cluster system may refer to jointly processing one task through a set of single computing units connected through a network. That is, each of the master node 100 and one or more subnodes 200 included in the cluster 1000 mentioned in this specification distributes and performs a specific task, for example, learning an artificial intelligence model, through a network. It is possible to mutually transmit and receive data for

The master node 100 and one or more sub-nodes 200 mentioned in this specification may form a horizontal relationship rather than a vertical relationship. For example, the master node 100 and one or more subnodes 200 may be peers having an equal relationship in a peer to peer (P2P) system.

That is, each of the master node 100 and one or more subnodes 200 may be changed depending on whether or not to perform an operation of dividing a specific task and allocating the divided task to other nodes. In other words, any one of the nodes included in the cluster 1000 may become a master node that divides a specific task and assigns the divided specific task to other nodes. However, it is not limited thereto.

Each of the master node 100 and one or more subnodes 200 may include any type of computer system or computer device, such as a microprocessor, mainframe computer, digital processor, portable device, and device controller. However, it is not limited thereto, and the master node 100 is, for example, a PC (personal computer), a notebook (note book), a mobile terminal, a smart phone (smart phone), a tablet PC (tablet pc) and It may include a wearable device and the like, and may include all types of terminals capable of accessing wired/wireless networks.

The master node 100 may include a processor 110, a communication unit 120 and a memory 130. However, the above-described components are not essential to implement the master node 100, the master node 100 may have more or less components than the components listed above. Here, each component may be configured as a separate chip, module, or device, or may be included in one device.

The communication unit 120 of the master node 100 may include a wired/wireless Internet module for network access. Wireless Internet technologies include wireless LAN (WLAN) (Wi-Fi), wireless broadband (Wibro), world interoperability for microwave access (Wimax), high speed downlink packet access (HSDPA), and the like. As a wired Internet technology, XDSL (Digital Subscriber Line), FTTH (Fibers to the home), PLC (Power Line Communication), and the like may be used. Also, the communication unit 120 may include a short range communication module. Bluetooth, Radio Frequency Identification (RFID), Infrared Data Association (IrDA), Ultra Wideband (UWB), ZigBee, and the like may be used as short-distance communication technologies.

The memory 130 may store any type of information generated or determined by the processor 110 and any type of information received by the communication unit 120 .

The memory 130 may include a memory and/or a permanent storage medium. Memory is a flash memory type, a hard disk type, a multimedia card micro type, a card type memory (eg SD or XD memory, etc.), RAM (Random Access Memory, RAM), SRAM (Static Random Access Memory), ROM (Read-Only Memory, ROM), EEPROM (Electrically Erasable Programmable Read-Only Memory), PROM (Programmable Read-Only Memory), magnetic memory, magnetic disk, optical disk At least one type of storage medium may be included.

On the other hand, the processor 110 of the master node 100 may process the overall operation of the master node 100 in general. The processor 110 may provide or process appropriate information or functions to the user by processing signals, data, information, etc. input or output through the components described above or by running an application program stored in the memory 130.

The processor 110 may be composed of one or more cores, and includes a central processing unit (CPU), a general purpose graphics processing unit (GPGPU), and a tensor processing unit (TPU) of the master node 100. : A tensor processing unit) may include any type of processor that executes instructions stored in the memory 130. The processor 110 may read a computer program stored in the memory 130 and perform operations according to some embodiments of the present disclosure.

As shown in FIG. 1 , one or more subnodes 200 may include a first subnode 210 and a second subnode 220 . However, since the above-described one or more subnodes 200 are expressed for convenience of description of the present disclosure, the one or more subnodes 200 may be more or less than the number of subnodes listed above.

Also, each of the one or more subnodes 200 may have similar or identical components to those included in the above-mentioned master node 100 .

Meanwhile, the client 300 may be related to a terminal of a user who wants to train an artificial intelligence model. For example, the client 300 is a personal computer (PC), a note book (note book), a mobile terminal (mobile terminal), a smart phone (smart phone), a tablet PC (tablet PC), and a wearable device (wearable device). It may include all types of terminals capable of accessing wired/wireless networks.

According to some embodiments of the present disclosure, the processor 110 of the master node 100 may receive a distributed resource request signal for model learning from the client 300 through the communication unit 120 . Also, the processor 110 may perform learning of the artificial intelligence model based on one or more subnodes 200 included in the cluster 1000 and the distributed resource request signal.

Hereinafter, a description of a specific method for the processor 110 of the master node 100 to perform training of an artificial intelligence model will be described later with reference to FIGS. 2 to 4 .

And, a description of the network structure of the artificial intelligence model learned by the master node 100 and/or one or more sub-nodes 200 will be described later with reference to FIG. 5 .

2 is a flowchart illustrating an example of a method for a master node to perform distributed learning of an artificial intelligence model according to some embodiments of the present disclosure.

Referring to FIG. 2 , the processor 110 of the master node 100 may receive a resource request signal for model learning from the client 300 through the communication unit 120 (S110). Here, the distributed resource request signal includes identification information capable of identifying the client 300, learning setting information required for learning the model, learning target information that is a learning goal of the model, and a job performed using the model to be learned ( job) may include at least one of the information.

Learning setting information required for model learning may include hyperparameter information related to the model. Here, the hierar parameter information includes learning rate, cost function, number of learning cycle iterations, weight initialization (for example, setting the range of weight values to be targeted for weight initialization), hidden unit number (eg, the number of hidden layers and the number of nodes in the hidden layer).

Meanwhile, as the processor 110 of the master node 100 receives the distributed resource request signal from the client 300, the master node 100 and one or more subnodes 200 included in one cluster 1000 Among them, a node to perform model learning may be determined (S120).

Specifically, the processor 110 may recognize a task for learning a model included in a distributed resource request signal. Also, the processor 110 may divide the task into sub-tasks and determine a node to perform the sub-task.

Hereinafter, a description of how the processor 110 of the master node 100 determines a node to perform a sub-task obtained by dividing a task for model learning will be described later with reference to FIG. 3 .

Additionally, the processor 110 may recognize whether learning setting information is included in the distributed resource request signal before determining a node to perform model learning. And, when recognizing that the distributed resource request signal includes the learning setting information, the processor 110 may determine the learning setting information based on the distributed resource request signal. That is, the processor 110 may recognize learning setting information included in the distributed resource request signal.

Meanwhile, when the processor 110 recognizes that learning setting information is not included in the distributed resource request signal, the processor 110 may recognize information about a job performed using a model included in the distributed resource request signal. Also, the processor 110 may determine learning setting information based on the information about the job.

Also, the processor 110 may determine a node to perform model learning among the master node 100 and one or more subnodes 200 based on the learning setting information.

According to some embodiments of the present disclosure, the processor 110 of the master node 100 may determine optimal learning setting information based on the distributed resource request signal.

Specifically, the processor 110 may recognize information about a job performed using a model included in a distributed resource request signal. Also, the processor 110 may determine optimal learning setting information based on information about the job of the model. Here, the memory 130 of the master node 100 may store optimal learning setting information corresponding to each of one or more jobs that can be performed by the artificial intelligence model. However, it is not limited thereto.

For example, the processor 110 may recognize that a model to be trained is a model that performs a job related to image recognition, based on the distributed resource request signal. In this case, the processor 110 may determine learning setting information with high efficiency for learning a model related to image recognition using optimal learning setting information corresponding to each of one or more jobs stored in the memory 130 .

For another example, the processor 110 may recognize that a model to be trained is a model that performs a job related to natural language processing, based on the distributed resource request signal. In this case, the processor 110 may determine learning setting information with high efficiency for learning of a model related to natural language processing using optimal learning setting information corresponding to each of one or more jobs stored in the memory 130 .

Meanwhile, the processor 110 of the master node 100 may control the communication unit 120 to transmit optimal learning setting information determined based on information about the job of the model to the client 300 .

Therefore, when the client 300 transmits the next distributed resource request signal for learning a model performing the same job to the master node 100, it can transmit optimal learning setting information together.

Referring back to FIG. 2 , the processor 110 of the master node 100 may obtain learning result information according to model learning. Then, the processor 110 may update the model based on the learning result information (S130).

Specifically, the processor 110 may obtain learning result information of a model performed in the master node 100 . Also, the processor 110 may receive learning result information of a model performed in one or more subnodes 200 through the communication unit 120 .

In addition, the processor 110 performs the above process every repetitive epoch from the master node 100 and each of the one or more subnodes 200, each time learning on a preset data set is completed, or each preset time interval. Learning result information can be obtained. However, it is not limited thereto.

Here, the learning result information may include at least one of version information of the learned model, a weight of the learned model, an amount of weight change according to the learning, and information about a learning method of the model.

The version information of the learned model may include information for identifying when the learned model was updated. For example, the version information includes information about the updated date, updated time, update subject, updated location (eg master node or one or more subnodes), meta information of the training data that caused the update, etc. can do. That is, when updating the model based on the learning result information, the processor 110 may also update the version information of the model.

The weight of the learned model may reflect knowledge extracted from a training data set used to train the model in the master node 100 and one or more subnodes 200 . Further, the weight of the learned model may be a weight for the entire learned model (ie, the learned model itself) or a weight for a part of the learned model.

Information about the learning method of the model may include a learning epoch, a learning rate per epoch, a required learning time, a bias, and the like. That is, the processor 110 may recognize how the model was learned by acquiring information on the learning method of the model.

Meanwhile, model learning may be performed independently in each of the master node 100 and one or more subnodes 200, or may be performed dependently in the master node 100 and one or more subnodes 200.

That is, the processor 110 may update the model by summing the learning result information obtained according to the model learning method (independent performance or dependent performance) or selecting any one learning result information.

According to some embodiments of the present disclosure, as described above, the processor 110 may update a model using learning result information.

Specifically, the processor 110 may transmit model A to one or more subnodes 200 . In addition, the processor 110 may receive model A′ learned in one or more subnodes 200 . In this case, the processor 110 may update some of the weights of model A using model A'. Here, the model A' received by the processor 110 from one or more subnodes 200 may be learning result information. Also, the learning result information may include a weight.

However, it is not limited to the above example, and the processor 110 may perform an update to change at least some of the weights of the model A to the weights of the model A'. Also, the processor 110 may update model A by replacing model A with model A'. Meanwhile, the processor 110 may evaluate the performance of the model A' using the test data set. When the performance of model A' obtained by inputting the test data set is superior to that of model A, the processor 110 may perform an update to replace model A with model A'. Meanwhile, if the performance of model A′ is lower than that of model A, the processor 110 may not perform an update on model A.

Information about a method of updating a model using the above-described learning result information is only a few examples of the present disclosure, and the present disclosure is not limited thereto.

Hereinafter, a description of how the processor 110 updates the model based on the learning result information will be described later with reference to FIG. 4 .

According to some embodiments of the present disclosure, the processor 110 of the master node 100 may determine optimal learning setting information based on learning result information.

Specifically, the processor 110 may determine the performance of a model trained in each of the master node 100 and one or more subnodes 200 . Also, the processor 110 may determine optimal learning setting information among learning setting information matched to each model based on performance of each model.

More specifically, the processor 110 performs a plurality of sub-learning result information according to learning of a model performed in each of the master node 100 and one or more sub-nodes 200 and sub-learning corresponding to the plurality of sub-learning result information. Setting information can be obtained. Also, the processor 110 may update the model to a plurality of sub-models using a plurality of sub-learning result information. Also, the processor 110 may determine the performance of each of the plurality of sub-models using the test data set.

The processor 110 may recognize first sub-learning result information used to update a specific model having the highest performance among a plurality of sub-models. Also, the processor 110 may determine first sub-learning setting information corresponding to the first sub-learning result information as optimal learning setting information.

Meanwhile, the processor 110 of the master node 100 may control the communication unit 120 to transmit optimal learning setting information determined based on the learning result information to the client 300 .

Therefore, when the client 300 transmits the next distributed resource request signal for learning a specific model to the master node 100, it can transmit optimal learning setting information together.

Referring back to FIG. 2 , the processor 110 may update a model based on learning result information and determine whether the updated model satisfies a preset condition.

Specifically, the processor 110 may obtain evaluation information of the updated model. Also, the processor 110 may determine whether a preset condition is satisfied based on the learning target information and the evaluation information included in the distributed resource request signal.

More specifically, the processor 110 may recognize that a preset condition is satisfied when the performance of the updated model included in the evaluation information is higher than the target performance of the model included in the learning target information.

Meanwhile, when the updated model satisfies a preset condition, the processor 110 may control the communication unit 120 to complete learning of the model and transmit the updated model to the client 300 (S140). .

Additionally, when the processor 110 transmits the updated model to the client 300, the above-described optimal learning setting information may be transmitted together.

On the other hand, if the updated model does not satisfy a preset condition, the processor 110 may re-learn the model. However, it is not limited thereto.

3 is a flowchart illustrating an example of a method in which a master node divides tasks for distributed learning of an artificial intelligence model according to some embodiments of the present disclosure. In the description of FIG. 3 , contents overlapping with those described in the descriptions of FIGS. 1 and 2 will not be described again, and will be described below focusing on the differences.

According to some embodiments of the present disclosure, the processor 110 of the master node 100 may receive a distributed resource request signal for model learning from the client 300 . In this case, the processor 110 may determine a node to perform model learning among the master node 100 and one or more subnodes 200 .

Specifically, referring to FIG. 3 , the processor 110 may recognize a task for learning a model included in a distributed resource request signal (S121). Also, the processor 110 may recognize resources of each of the one or more subnodes 200 in order to distribute the tasks. Here, the resource may be a computing resource. For example, the resource may be at least one of CPU, GPU, memory, and disk. In addition, the task (ie, model learning) is independently performed in each of the master node 100 and one or more subnodes 200, or performed dependently in the master node 100 and one or more subnodes 200. It can be.

Meanwhile, the processor 110 may divide the task into sub-tasks based on resources of each of the one or more sub-nodes 200 (S123).

Specifically, the processor 110 may determine one or more specific nodes to which a sub task is to be assigned based on resources of each of the one or more recognized sub nodes 200 . And, for example, the processor 110 may divide the task into sub-tasks to correspond to the size of each resource of one or more specific nodes. More specifically, the processor 110 divides and allocates (ie, load balancing) sub-tasks so that loads are evenly distributed among nodes operating in parallel within the cluster 1000 . However, the present invention is not limited thereto, and the processor 110 may equally divide the size of the sub-task based on the resources of each of the one or more recognized sub-nodes 200 .

Then, the processor 110 may determine a node to perform the sub task (S124). That is, the processor 110 may control the communication unit 120 to transmit the sub-task to a node to perform the sub-task.

On the other hand, as described above, the task (ie, model learning) is independently performed in each of the master node 100 and one or more subnodes 200, or the master node 100 and one or more subnodes 200 can be performed dependently. For example, a task may be configured to be performed in parallel or asynchronously on one or more nodes, or serially or synchronously on one or more nodes. For example, in order to save time and resources required for learning, learning for each subset of learning data obtained by dividing the learning data set based on an arbitrary criterion may constitute a sub-task. Also, for another example, a sub-task may be configured so that each node performs learning on a training data set for the performance of the learned model.

First, when a task is independently performed in each of the master node 100 and one or more subnodes 200, the size of the task and the subtask may be the same. For example, if a task has a size of 100, a sub-task may also have a size of 100.

In other words, each of the master node 100 and one or more subnodes 200 may process subtasks having the same size as the size of the task in parallel. In this case, learning setting information for each sub-task may be different.

That is, each of the master node 100 and one or more subnodes 200 may perform model learning through different methods. For example, a model learned in each of the master node 100, the first sub-node 210, and the second sub-node 220 may be the same as model A. In addition, each of the master node 100, the first sub-node 210, and the second sub-node 220 may perform learning of model A using different learning methods (learning setting information). Here, the sizes of tasks for learning model A in each of the master node 100, the first subnode 210, and the second subnode 220 may be the same. However, it is not limited thereto, and the size of the task may be different depending on the learning method.

In this case, the processor 110 may select one of a plurality of sub-learning result information according to model learning performed in the master node 100 and each of the one or more sub-nodes 200 . And, the processor 110 may update the model using the selected learning result information.

Meanwhile, when a task is performed dependently on the master node 100 and one or more sub-nodes 200, the sum of the sizes of the sub-tasks may be equal to the size of the tasks. For example, if a task has a size of 100, there may be 10 sub-tasks having a size of 10 (in this case, the number of master node 100 and one or more sub-nodes 200 is 10). For another example, if a task has a size of 100, there may be 2 sub-tasks with a size of 20 and 6 sub-tasks with a size of 10 (in this case, the master node 100 and one or more sub-nodes). (the number of 200 is 5).

In other words, each of the master node 100 and one or more sub-nodes 200 may process tasks by dividing them into sub-tasks. In this case, learning setting information for each sub-task may be the same. For example, the model learned in the master node 100 may be model A1, the model learned in the first subnode 210 may be model A2, and the model learned in the second subnode 220 may be model A3. . In addition, each of the master node 100, the first sub-node 210, and the second sub-node 220 performs learning of Model A1, Model A2, and Model A3 using the same learning method (learning setting information). can Here, the sizes of tasks for learning each of the model A1, model A2, and model A3 in the master node 100, the first sub-node 210, and the second sub-node 220 may be the same or different.

In this case, the processor 110 may sum up a plurality of sub-learning result information according to model learning performed by the master node 100 and each of the one or more sub-nodes 200 . And, the processor 110 may update the model using the summed learning result information.

Hereinafter, as the task is independently performed in each of the master node 100 and one or more subnodes 200 or dependently performed in the master node 100 and one or more subnodes 200, updating the model A description of the method will be described later with reference to FIG. 4 .

4 is a flowchart illustrating an example of a method for a master node to update an artificial intelligence model based on learning result information according to some embodiments of the present disclosure. In the description of FIG. 4 , contents overlapping with those described above in the description of FIGS. 1 to 3 will not be described again, and will be described below focusing on contents having differences.

According to some embodiments of the present disclosure, the task (ie, model learning) is independently performed in each of the master node 100 and one or more subnodes 200, or the master node 100 and one or more subnodes ( 200) can be performed dependently.

Meanwhile, as the task is performed independently or dependently, a method of updating a model using acquired learning result information may be different.

Specifically, referring to FIG. 4 , the processor 110 may recognize whether model learning is dependently performed in the master node 100 and one or more subnodes 200 (S131).

More specifically, the processor 110 recognizes whether or not learning of the model has been performed dependently on the master node 100 and one or more sub-nodes 200 based on how the task is divided into sub-tasks. can do.

And, when the processor 110 recognizes that the learning of the model is dependently performed in the master node 100 and one or more sub-nodes 200 (S131, Yes), it is performed in each of the one or more sub-nodes 200. First learning result information according to learning of the model may be received through the communication unit 120 (S132). In addition, the processor 110 may obtain second learning result information according to model learning performed by the master node 100 (S133).

And, the processor 110 may obtain learning result information by summing the first learning result information and the second learning result information. Then, the processor 110 may update the model based on the learning result information (S134).

On the other hand, when the processor 110 recognizes that the learning of the model is not dependently performed by the master node 100 and one or more subnodes 200 (S131, No), the learning of the model is performed by the master node 100 and It can be recognized that each of the one or more sub-nodes 200 is independently performed.

In this case, the processor 110 may obtain a plurality of sub-learning result information according to model learning performed in each of the master node 100 and one or more sub-nodes 200 (S135). In addition, the processor 110 may update a model to a plurality of sub-models using a plurality of sub-learning result information. Also, the processor 110 may determine the performance of each of the plurality of sub-models using the test data set (S136).

Also, the processor 110 may determine learning result information of a model determined based on the performance of each of a plurality of sub-models as learning result information. Also, the processor 110 may determine the model determined based on performance as an updated model (S137).

As described above, the processor 110 according to some embodiments of the present disclosure determines learning result information using a different method according to the method (independent or dependent) in which model learning is performed, and the determined learning result information. can be used to update the model.

5 is an exemplary diagram briefly illustrating a network structure of artificial intelligence models that can be learned according to some embodiments of the present disclosure.

Throughout this specification, computational model, neural network, network function, and neural network may be used interchangeably. A neural network may consist of a set of interconnected computational units, which may generally be referred to as nodes. These nodes may also be referred to as neurons. A neural network includes one or more nodes. Nodes (or neurons) constituting neural networks may be interconnected by one or more links.

In a neural network, one or more nodes connected through a link may form a relative relationship of an input node and an output node. The concept of an input node and an output node is relative, and any node in an output node relationship with one node may have an input node relationship with another node, and vice versa. As described above, an input node to output node relationship may be created around a link. More than one output node can be connected to one input node through a link, and vice versa.

In a relationship between an input node and an output node connected through one link, the value of the output node may be determined based on data input to the input node. Here, a node interconnecting an input node and an output node may have a weight. The weight may be variable, and may be changed by a user or an algorithm in order to perform a function desired by the neural network. For example, when one or more input nodes are interconnected by respective links to one output node, the output node is set to a link corresponding to values input to input nodes connected to the output node and respective input nodes. An output node value may be determined based on the weight.

As described above, in the neural network, one or more nodes are interconnected through one or more links to form an input node and output node relationship in the neural network. Characteristics of the neural network may be determined according to the number of nodes and links in the neural network, an association between the nodes and links, and a weight value assigned to each link. For example, when there are two neural networks having the same number of nodes and links and different weight values between the links, the two neural networks may be recognized as different from each other.

A neural network may include one or more nodes. Some of the nodes constituting the neural network may form one layer based on distances from the first input node. For example, a set of nodes having a distance of n from the first input node may constitute n layers. The distance from the first input node may be defined by the minimum number of links that must be passed through to reach the corresponding node from the first input node. However, the definition of such a layer is arbitrary for explanation, and the order of a layer in a neural network may be defined in a method different from the above. For example, a layer of nodes may be defined by a distance from a final output node.

An initial input node may refer to one or more nodes to which data is directly input without going through a link in relation to other nodes among nodes in the neural network. Alternatively, in a relationship between nodes based on a link in a neural network, it may refer to nodes that other input nodes connected to a link do not have. Similarly, the final output node may refer to one or more nodes that do not have an output node in relation to other nodes among nodes in the neural network. Also, the hidden node may refer to nodes constituting the neural network other than the first input node and the last output node. In the neural network according to some embodiments of the present disclosure, the number of nodes in the input layer may be the same as the number of nodes in the output layer, and the number of nodes decreases and then increases again as the number of nodes progresses from the input layer to the hidden layer. can In addition, the neural network according to some other embodiments of the present disclosure may be a neural network in which the number of nodes of the input layer may be less than the number of nodes of the output layer, and the number of nodes decreases as the number of nodes increases from the input layer to the hidden layer. have. In addition, the neural network according to some other embodiments of the present disclosure is a neural network in which the number of nodes in the input layer may be greater than the number of nodes in the output layer, and the number of nodes increases as the number of nodes progresses from the input layer to the hidden layer. can A neural network according to some other embodiments of the present disclosure may be a neural network in the form of a combination of the aforementioned neural networks.

A deep neural network (DNN) may refer to a neural network including a plurality of hidden layers in addition to an input layer and an output layer. Deep neural networks can reveal latent structures in data. In other words, it can identify the latent structure of a photo, text, video, sound, or music (e.g., what objects are in the photo, what the content and emotion of the text are, what the content and emotion of the audio are, etc.). . Deep neural networks include convolutional neural networks (CNN), recurrent neural networks (RNNs), auto encoders, generative adversarial networks (GANs), and restricted Boltzmann machines (RBMs). boltzmann machine), deep belief network (DBN), Q network, U network, Siamese network, and the like. The description of the deep neural network described above is only an example, and the present disclosure is not limited thereto.

In some embodiments of this disclosure, the network function may include an auto encoder. An autoencoder may be a type of artificial neural network for outputting output data similar to input data. An auto-encoder may include at least one hidden layer, and an odd number of hidden layers may be disposed between input and output layers. The number of nodes of each layer may be reduced from the number of nodes of the input layer to an intermediate layer called the bottleneck layer (encoding), and then expanded symmetrically with the reduction from the bottleneck layer to the output layer (symmetrical to the input layer). In this case, the dimension reduction layer and the dimension restoration layer may be symmetric, but the present disclosure is not limited thereto, and nodes of the dimension reduction layer and the dimension restoration layer may not be symmetric. Autoencoders can perform non-linear dimensionality reduction. The number of input layers and output layers may correspond to the number of remaining sensors after preprocessing of input data. In the auto-encoder structure, the number of hidden layer nodes included in the encoder may decrease as the distance from the input layer increases. If the number of nodes in the bottleneck layer (the layer with the fewest nodes located between the encoder and decoder) is too small, a sufficient amount of information may not be conveyed, so more than a certain number (e.g., more than half of the input layer, etc.) ) may be maintained.

The neural network may be trained using at least one of supervised learning, unsupervised learning, and semi-supervised learning. The learning of neural networks is to minimize errors in the output. In the learning of the neural network, the learning data is repeatedly input into the neural network, the output of the neural network for the training data and the error of the target are calculated, and the error of the neural network is transferred from the output layer of the neural network to the input layer in the direction of reducing the error. It is a process of updating the weight of each node of the neural network by backpropagating in the same direction. In the case of teacher learning, the learning data in which the correct answer is labeled is used for each learning data (ie, the labeled learning data), and in the case of comparative teacher learning, the correct answer may not be labeled in each learning data. That is, for example, learning data in the case of teacher learning regarding data classification may be data in which each learning data is labeled with a category. Labeled training data is input to a neural network, and an error may be calculated by comparing an output (category) of the neural network and a label of the training data. As another example, in the case of comparative history learning for data classification, an error may be calculated by comparing input learning data with a neural network output. The calculated error is back-propagated in a reverse direction (ie, from the output layer to the input layer) in the neural network, and the connection weight of each node of each layer of the neural network may be updated according to the back-propagation. The amount of change in the connection weight of each updated node may be determined according to a learning rate. The neural network's computation of input data and backpropagation of errors can constitute a learning cycle (epoch). The learning rate may be applied differently according to the number of iterations of the learning cycle of the neural network. For example, a high learning rate may be used in the early stage of neural network training to increase efficiency by allowing the neural network to quickly obtain a certain level of performance, and a low learning rate may be used in the late stage to increase accuracy.

In neural network learning, generally, training data can be a subset of real data (ie, data to be processed using the trained neural network), and therefore, errors for training data are reduced, but errors for real data are reduced. There may be incremental learning cycles. Overfitting is a phenomenon in which errors for actual data increase due to excessive learning on training data. For example, a phenomenon in which a neural network that has learned a cat by showing a yellow cat does not recognize that it is a cat when it sees a cat other than yellow may be a type of overfitting. Overfitting can act as a cause of increasing the error of machine learning algorithms. Various optimization methods can be used to prevent such overfitting. In order to prevent overfitting, methods such as increasing training data, regularization, and omitting some nodes of a network in the process of learning may be applied.

A computer readable medium storing a data structure according to some embodiments of the present disclosure is disclosed.

Data structure can refer to the organization, management, and storage of data that enables efficient access and modification of data. Data structure may refer to the organization of data to solve a specific problem (eg, data retrieval, data storage, data modification in the shortest time). A data structure may be defined as a physical or logical relationship between data elements designed to support a specific data processing function. A logical relationship between data elements may include a connection relationship between data elements that a user thinks. A physical relationship between data elements may include an actual relationship between data elements physically stored on a computer-readable storage medium (eg, a hard disk). The data structure may specifically include a set of data, a relationship between data, and a function or command applicable to the data. Through an effectively designed data structure, a computing device can perform calculations while using minimal resources of the computing device. Specifically, the computing device can increase the efficiency of operation, reading, insertion, deletion, comparison, exchange, and search through an effectively designed data structure.

The data structure can be divided into a linear data structure and a non-linear data structure according to the shape of the data structure. A linear data structure may be a structure in which only one data is connected after one data. Linear data structures may include lists, stacks, queues, and decks. A list may refer to a series of data sets in which order exists internally. The list may include a linked list. A linked list may be a data structure in which data are connected in such a way that each data is connected in a line with a pointer. In a linked list, a pointer can contain information about connection to the next or previous data. A linked list can be expressed as a singly linked list, a doubly linked list, or a circular linked list depending on the form. A stack can be a data enumeration structure that allows limited access to data. A stack can be a linear data structure in which data can be processed (eg, inserted or deleted) at only one end of the data structure. The data stored in the stack may be a LIFO-Last in First Out (Last in First Out) data structure. A queue is a data listing structure that allows limited access to data, and unlike a stack, it can be a data structure (FIFO-First in First Out) in which data stored later comes out later. A deck can be a data structure that can handle data from either end of the data structure.

The nonlinear data structure may be a structure in which a plurality of data are connected after one data. The non-linear data structure may include a graph data structure. A graph data structure can be defined as a vertex and an edge, and an edge can include a line connecting two different vertices. A graph data structure may include a tree data structure. The tree data structure may be a data structure in which one path connects two different vertices among a plurality of vertices included in the tree. That is, it may be a data structure that does not form a loop in a graph data structure.

Throughout this specification, computational model, neural network, network function, and neural network may be used interchangeably. (Hereinafter, it is unified and described as a neural network.) The data structure may include a neural network. And the data structure including the neural network may be stored in a computer readable medium. The data structure including the neural network may also include data input to the neural network, weights of the neural network, hyperparameters of the neural network, data obtained from the neural network, an activation function associated with each node or layer of the neural network, and a loss function for learning the neural network. have. A data structure including a neural network may include any of the components described above. In other words, the data structure including the neural network includes all or all of the data input to the neural network, weights of the neural network, hyperparameters of the neural network, data obtained from the neural network, activation function associated with each node or layer of the neural network, and loss function for training the neural network. It may be configured to include any combination of. In addition to the foregoing configurations, the data structure comprising the neural network may include any other information that determines the characteristics of the neural network. In addition, the data structure may include all types of data used or generated in the computational process of the neural network, but is not limited to the above. A computer readable medium may include a computer readable recording medium and/or a computer readable transmission medium. A neural network may consist of a set of interconnected computational units, which may generally be referred to as nodes. These nodes may also be referred to as neurons. A neural network includes one or more nodes.

The data structure may include data input to the neural network. A data structure including data input to the neural network may be stored in a computer readable medium. Data input to the neural network may include training data input during a neural network learning process and/or input data input to a neural network that has been trained. Data input to the neural network may include pre-processed data and/or data subject to pre-processing. Pre-processing may include a data processing process for inputting data to a neural network. Accordingly, the data structure may include data subject to pre-processing and data generated by pre-processing. The foregoing data structure is only an example, and the present disclosure is not limited thereto.

The data structure may include data input to or output from the neural network. A data structure including data input or output to the neural network may be stored in a computer readable medium. A data structure stored in a computer readable medium may include data input during a neural network inference process or output data output as a result of neural network inference. Also, since the data structure may include data processed by a specific data processing method, it may include data before and after processing. Accordingly, the data structure may include data subject to processing and data processed through a data processing method.

The data structure may include the weights of the neural network. (In this specification, weights and parameters may be used in the same meaning.) Also, a data structure including weights of a neural network may be stored in a computer readable medium. A neural network may include a plurality of weights. The weight may be variable, and may be changed by a user or an algorithm in order to perform a function desired by the neural network. For example, when one or more input nodes are interconnected by respective links to one output node, the output node is set to a link corresponding to values input to input nodes connected to the output node and respective input nodes. An output node value can be determined based on the parameter. The foregoing data structure is only an example, and the present disclosure is not limited thereto.

As a non-limiting example, the weights may include weights that are varied during neural network training and/or weights for which neural network training has been completed. The variable weight in the neural network learning process may include a weight at the time the learning cycle starts and/or a variable weight during the learning cycle. The weights for which neural network learning has been completed may include weights for which learning cycles have been completed. Accordingly, the data structure including the weights of the neural network may include a data structure including weights that are variable during the neural network learning process and/or weights for which neural network learning is completed. Therefore, it is assumed that the above-described weights and/or combinations of weights are included in the data structure including the weights of the neural network. The foregoing data structure is only an example, and the present disclosure is not limited thereto.

The data structure including the weights of the neural network may be stored in a computer readable storage medium (eg, a memory or a hard disk) after going through a serialization process. Serialization can be the process of converting a data structure into a form that can be stored on the same or another computing device and later reconstructed and used. A computing device may serialize data structures to transmit and receive data over a network. The data structure including the weights of the serialized neural network may be reconstructed on the same computing device or another computing device through deserialization. The data structure including the weights of the neural network is not limited to serialization. Furthermore, the data structure including the weights of the neural network is a data structure for increasing the efficiency of operation while minimizing the resource of the computing device (for example, B-Tree, Trie, m-way search tree, AVL tree, Red-Black Tree). The foregoing is only an example, and the present disclosure is not limited thereto.

The data structure may include hyper-parameters of the neural network. Also, the data structure including the hyperparameters of the neural network may be stored in a computer readable medium. A hyperparameter may be a variable variable by a user. Hyperparameters include, for example, learning rate, cost function, number of learning cycle iterations, weight initialization (eg, setting the range of weight values to be targeted for weight initialization), hidden unit number (eg, the number of hidden layers and the number of nodes in the hidden layer). The foregoing data structure is only an example, and the present disclosure is not limited thereto.

6 shows a general schematic diagram of an example computing environment in which embodiments of the present disclosure may be implemented.

The computer 1102 shown in FIG. 6 may correspond to each of the master node 100 and/or one or more subnodes 200 included in the cluster 1000 .

Although the present disclosure has generally been described above in terms of computer-executable instructions that can be executed on one or more computers, those skilled in the art will appreciate that the present disclosure may be implemented in combination with other program modules and/or as a combination of hardware and software. will know

Generally, modules herein include routines, procedures, programs, components, data structures, etc. that perform particular tasks or implement particular abstract data types. It will also be appreciated by those skilled in the art that the methods of the present disclosure may be used in single-processor or multiprocessor computer systems, minicomputers, mainframe computers, as well as personal computers, handheld computing devices, microprocessor-based or programmable consumer electronics, and the like (each of which is It will be appreciated that other computer system configurations may be implemented, including those that may be operative in connection with one or more associated devices.

The described embodiments of the present disclosure may also be practiced in distributed computing environments where certain tasks are performed by remote processing devices that are linked through a communications network. In a distributed computing environment, program modules may be located in both local and remote memory storage devices.

A computer typically includes a variety of computer readable media. Media accessible by a computer includes volatile and nonvolatile media, transitory and non-transitory media, removable and non-removable media. By way of example, and not limitation, computer readable media may include computer readable storage media and computer readable transmission media.

Computer readable storage media are volatile and nonvolatile media, transitory and non-transitory, removable and non-removable media implemented in any method or technology for storage of information such as computer readable instructions, data structures, program modules or other data. includes media Computer readable storage media may include RAM, ROM, EEPROM, flash memory or other memory technology, CD-ROM, digital video disk (DVD) or other optical disk storage device, magnetic cassette, magnetic tape, magnetic disk storage device or other magnetic storage device. device, or any other medium that can be accessed by a computer and used to store desired information.

A computer readable transmission medium typically embodies computer readable instructions, data structures, program modules or other data in a modulated data signal such as a carrier wave or other transport mechanism. Including all information delivery media. The term modulated data signal means a signal that has one or more of its characteristics set or changed so as to encode information within the signal. By way of example, and not limitation, computer readable transmission media includes wired media such as a wired network or direct-wired connection, and wireless media such as acoustic, RF, infrared, and other wireless media. Combinations of any of the above are also intended to be included within the scope of computer readable transmission media.

An exemplary environment 1100 implementing various aspects of the present disclosure is shown including a computer 1102, which includes a processing unit 1104, a system memory 1106, and a system bus 1108. do. System bus 1108 couples system components, including but not limited to system memory 1106 , to processing unit 1104 . Processing unit 1104 may be any of a variety of commercially available processors. Dual processor and other multiprocessor architectures may also be used as the processing unit 1104.

System bus 1108 may be any of several types of bus structures that may additionally be interconnected to a memory bus, a peripheral bus, and a local bus using any of a variety of commercial bus architectures. System memory 1106 includes read only memory (ROM) 1110 and random access memory (RAM) 1112 . A basic input/output system (BIOS) is stored in non-volatile memory 1110, such as ROM, EPROM, or EEPROM, and is a basic set of information that helps transfer information between components within computer 1102, such as during startup. contains routines. RAM 1112 may also include high-speed RAM, such as static RAM, for caching data.

The computer 1102 may also include an internal hard disk drive (HDD) 1114 (eg, EIDE, SATA) - the internal hard disk drive 1114 may also be configured for external use within a suitable chassis (not shown). Yes—a magnetic floppy disk drive (FDD) 1116 (e.g., for reading from or writing to removable diskette 1118), and an optical disk drive 1120 (e.g., CD-ROM) for reading disc 1122 or reading from or writing to other high capacity optical media such as DVDs). The hard disk drive 1114, magnetic disk drive 1116, and optical disk drive 1120 are connected to the system bus 1108 by a hard disk drive interface 1124, magnetic disk drive interface 1126, and optical drive interface 1128, respectively. ) can be connected to The interface 1124 for external drive implementation includes, for example, at least one or both of Universal Serial Bus (USB) and IEEE 1394 interface technologies.

These drives and their associated computer readable media provide non-volatile storage of data, data structures, computer executable instructions, and the like. In the case of computer 1102, drives and media correspond to storing any data in a suitable digital format. Although the description of computer-readable storage media above refers to HDDs, removable magnetic disks, and removable optical media such as CDs or DVDs, those skilled in the art can use zip drives, magnetic cassettes, flash memory cards, cartridges, It will be appreciated that other types of storage media readable by the computer may also be used in the exemplary operating environment and any such media may include computer executable instructions for performing the methods of the present disclosure. .

A number of program modules may be stored on the drive and RAM 1112, including an operating system 1130, one or more application programs 1132, other program modules 1134, and program data 1136. All or portions of the operating system, applications, modules and/or data may also be cached in RAM 1112. It will be appreciated that the present disclosure may be implemented in a variety of commercially available operating systems or combinations of operating systems.

A user may enter commands and information into the computer 1102 through one or more wired/wireless input devices, such as a keyboard 1138 and a pointing device such as a mouse 1140. Other input devices (not shown) may include a microphone, IR remote control, joystick, game pad, stylus pen, touch screen, and the like. Although these and other input devices are often connected to the processing unit 1104 through an input device interface 1142 that is connected to the system bus 1108, a parallel port, IEEE 1394 serial port, game port, USB port, IR interface, may be connected by other interfaces such as the like.

A monitor 1144 or other type of display device is also connected to the system bus 1108 through an interface such as a video adapter 1146. In addition to the monitor 1144, computers typically include other peripheral output devices (not shown) such as speakers, printers, and the like.

Computer 1102 may operate in a networked environment using logical connections to one or more remote computers, such as remote computer(s) 1148 via wired and/or wireless communications. Remote computer(s) 1148 may be a workstation, server computer, router, personal computer, handheld computer, microprocessor-based entertainment device, peer device, or other common network node, and may generally refer to computer 1102. Although many or all of the described components are included, for brevity, only memory storage device 1150 is shown. The logical connections shown include wired/wireless connections to a local area network (LAN) 1152 and/or a larger network, such as a wide area network (WAN) 1154 . Such LAN and WAN networking environments are common in offices and corporations and facilitate enterprise-wide computer networks, such as intranets, all of which can be connected to worldwide computer networks, such as the Internet.

When used in a LAN networking environment, computer 1102 connects to local network 1152 through wired and/or wireless communication network interfaces or adapters 1156. Adapter 1156 may facilitate wired or wireless communications to LAN 1152, which also includes a wireless access point installed therein to communicate with wireless adapter 1156. When used in a WAN networking environment, computer 1102 may include a modem 1158, be connected to a communications server on WAN 1154, or other device that establishes communications over WAN 1154, such as over the Internet. have the means A modem 1158, which may be internal or external and a wired or wireless device, is connected to the system bus 1108 through a serial port interface 1142. In a networked environment, program modules described for computer 1102, or portions thereof, may be stored on remote memory/storage device 1150. It will be appreciated that the network connections shown are exemplary and other means of establishing a communication link between computers may be used.

Computer 1102 is any wireless device or entity that is deployed and operating in wireless communication, eg, printers, scanners, desktop and/or portable computers, portable data assistants (PDAs), communication satellites, wireless detectable tags associated with It operates to communicate with arbitrary equipment or places and telephones. This includes at least Wi-Fi and Bluetooth wireless technologies. Thus, the communication may be a predefined structure as in conventional networks or simply an ad hoc communication between at least two devices.

Wi-Fi (Wireless Fidelity) makes it possible to connect to the Internet without wires. Wi-Fi is a wireless technology, such as a cell phone, that allows such devices, eg, computers, to transmit and receive data both indoors and outdoors, i.e. anywhere within coverage of a base station. Wi-Fi networks use a radio technology called IEEE 802.11 (a,b,g, etc.) to provide secure, reliable, and high-speed wireless connections. Wi-Fi can be used to connect computers to each other, to the Internet, and to wired networks (using IEEE 802.3 or Ethernet). Wi-Fi networks can operate in the unlicensed 2.4 and 5 GHz radio bands, for example, at 11 Mbps (802.11a) or 54 Mbps (802.11b) data rates, or on products that include both bands (dual band). have.

Those skilled in the art will understand that the various illustrative logical blocks, modules, processors, means, circuits, and algorithm steps described in connection with the embodiments disclosed herein are electronic hardware, (for convenience) , may be implemented by various forms of program or design code (referred to herein as “software”) or a combination of both. To clearly illustrate this interchangeability of hardware and software, various illustrative components, blocks, modules, circuits, and steps have been described above generally in terms of their functionality. Whether such functionality is implemented as hardware or software depends on the particular application and the design constraints imposed on the overall system. Skilled artisans may implement the described functionality in varying ways for each particular application, but such implementation decisions should not be interpreted as causing a departure from the scope of the present disclosure.

Various embodiments presented herein may be implemented as a method, apparatus, or article of manufacture using standard programming and/or engineering techniques. The term "article of manufacture" includes a computer program or media accessible from any computer-readable device. For example, computer-readable storage media include magnetic storage devices (eg, hard disks, floppy disks, magnetic strips, etc.), optical disks (eg, CDs, DVDs, etc.), smart cards, and flash memory devices (eg, EEPROM, cards, sticks, key drives, etc.), but are not limited thereto. The term “machine-readable medium” includes, but is not limited to, wireless channels and various other media that can store, hold, and/or convey instruction(s) and/or data.

The description of the presented embodiments is provided to enable any person skilled in the art to make or use the present disclosure. Various modifications to these embodiments will be apparent to those skilled in the art of this disclosure, and the general principles defined herein may be applied to other embodiments without departing from the scope of this disclosure. Thus, the present disclosure is not to be limited to the embodiments presented herein, but is to be interpreted in the widest scope consistent with the principles and novel features presented herein.

Claims (17)

A computer program stored on a computer readable storage medium, the computer program including instructions for causing a processor of a master node to perform the following steps, wherein the steps are:
Receiving a distributed resource request signal for model learning from a client;
determining a node to perform learning of the model among the master node and one or more subnodes included in one cluster, upon receiving the distributed resource request signal;
obtaining learning result information according to learning of the model, and updating the model based on the learning result information; and
completing learning of the model and transmitting the updated model to the client when the updated model satisfies a preset condition;
Including,
Upon receiving the distributed resource request signal, determining a node to perform learning of the model among the master node and one or more subnodes included in one cluster,
Recognizing a task for learning the model included in the distributed resource request signal;
Recognizing a resource of each of the one or more subnodes to distribute the task;
Dividing the task into subtasks based on resources of each of the one or more subnodes; and
determining a node to perform the sub-task;
including,
A computer program stored on a computer readable storage medium.
delete According to claim 1,
The task is independently performed in each of the master node and the one or more subnodes, or performed dependently in the master node and the one or more subnodes.
A computer program stored on a computer readable storage medium.
A computer program stored on a computer readable storage medium, the computer program including instructions for causing a processor of a master node to perform the following steps, wherein the steps are:
Receiving a distributed resource request signal for model learning from a client;
determining a node to perform learning of the model among the master node and one or more subnodes included in one cluster, upon receiving the distributed resource request signal;
obtaining learning result information according to learning of the model, and updating the model based on the learning result information; and
completing learning of the model and transmitting the updated model to the client when the updated model satisfies a preset condition;
Including,
Upon receiving the distributed resource request signal, determining a node to perform learning of the model among the master node and one or more subnodes included in one cluster,
determining learning setting information based on the distributed resource request signal; and
determining a node to perform learning of the model among the master node and one or more subnodes based on the learning setting information;
including,
A computer program stored on a computer readable storage medium.
According to claim 1,
recognizing information about a job performed using the model included in the distributed resource request signal; and
determining optimal learning setting information based on the information about the job;
Including more,
A computer program stored on a computer readable storage medium.
According to claim 1,
Obtaining learning result information according to learning of the model and updating the model based on the learning result information,
determining optimal learning setting information based on the learning result information;
including,
A computer program stored on a computer readable storage medium.
According to claim 6,
The step of determining optimal learning setting information based on the learning result information,
Determining the performance of the model learned in each of the master node and the one or more subnodes, and determining optimal learning setting information among learning setting information matched to each model based on the performance of each model;
including,
A computer program stored on a computer readable storage medium.
A computer program stored on a computer readable storage medium, the computer program including instructions for causing a processor of a master node to perform the following steps, wherein the steps are:
Receiving a distributed resource request signal for model learning from a client;
determining a node to perform learning of the model among the master node and one or more subnodes included in one cluster, upon receiving the distributed resource request signal;
obtaining learning result information according to learning of the model, and updating the model based on the learning result information; and
completing learning of the model and transmitting the updated model to the client when the updated model satisfies a preset condition;
including
Acquiring learning result information according to learning of the model, and updating the model based on the learning result information includes determining optimal learning setting information based on the learning result information,
Determining optimal learning setting information based on the learning result information may include determining the performance of a model learned in the master node and each of the one or more subnodes, and determining the performance of each model based on the performance of each model. Determining optimal learning setting information among learning setting information matched to
Determining the performance of the model learned in the master node and each of the one or more subnodes, and determining optimal learning setting information among learning setting information matched to each model based on the performance of each model,
obtaining a plurality of sub-learning result information according to the model learning performed in the master node and each of the one or more sub-nodes and sub-learning setting information corresponding to the plurality of sub-learning result information;
updating the model into a plurality of sub-models using the plurality of sub-learning result information, and determining performance of each of the plurality of sub-models using a test data set;
recognizing first sub-learning result information used to update a specific model having the highest performance among the plurality of sub-models; and
determining first sub-learning setting information corresponding to the first sub-learning result information as the optimal learning setting information;
including,
A computer program stored on a computer readable storage medium.
A computer program stored on a computer readable storage medium, the computer program including instructions for causing a processor of a master node to perform the following steps, wherein the steps are:
Receiving a distributed resource request signal for model learning from a client;
determining a node to perform learning of the model among the master node and one or more subnodes included in one cluster, upon receiving the distributed resource request signal;
obtaining learning result information according to learning of the model, and updating the model based on the learning result information; and
completing learning of the model and transmitting the updated model to the client when the updated model satisfies a preset condition;
Including,
Obtaining learning result information according to learning of the model and updating the model based on the learning result information,
receiving first learning result information according to learning of the model performed in each of the one or more sub-nodes;
obtaining second learning result information according to learning of the model performed in the master node; and
acquiring the learning result information by summing the first learning result information and the second learning result information, and updating the model based on the learning result information;
including,
A computer program stored on a computer readable storage medium.
A computer program stored on a computer readable storage medium, the computer program including instructions for causing a processor of a master node to perform the following steps, wherein the steps are:
Receiving a distributed resource request signal for model learning from a client;
determining a node to perform learning of the model among the master node and one or more subnodes included in one cluster, upon receiving the distributed resource request signal;
obtaining learning result information according to learning of the model, and updating the model based on the learning result information; and
completing learning of the model and transmitting the updated model to the client when the updated model satisfies a preset condition;
Including,
Obtaining learning result information according to learning of the model and updating the model based on the learning result information,
obtaining a plurality of sub-learning result information according to the learning of the model performed in the master node and each of the one or more sub-nodes;
updating the model into a plurality of sub-models using the plurality of sub-learning result information, and determining performance of each of the plurality of sub-models using a test data set; and
Based on the performance of each of the plurality of sub-models, learning result information of a model selected from among the plurality of sub-models is determined as the learning result information, and the model selected from among the plurality of sub-models based on the performance is determined as the updated model. determining as a model;
including,
A computer program stored on a computer readable storage medium.
A computer program stored on a computer readable storage medium, the computer program including instructions for causing a processor of a master node to perform the following steps, wherein the steps are:
Receiving a distributed resource request signal for model learning from a client;
determining a node to perform learning of the model among the master node and one or more subnodes included in one cluster, upon receiving the distributed resource request signal;
obtaining learning result information according to learning of the model, and updating the model based on the learning result information; and
completing learning of the model and transmitting the updated model to the client when the updated model satisfies a preset condition;
Including,
When the updated model satisfies a preset condition, completing the learning of the model and transmitting the updated model to the client,
obtaining evaluation information of the updated model;
determining whether the predetermined condition is satisfied based on the learning target information and the evaluation information included in the distributed resource request signal;
including,
A computer program stored on a computer readable storage medium.
According to claim 11,
Determining whether or not the preset condition is satisfied based on the learning target information and the evaluation information included in the distributed resource request signal,
recognizing that the preset condition is satisfied when the performance of the updated model included in the evaluation information is higher than the target performance of the model included in the learning target information;
including,
A computer program stored on a computer readable storage medium.
A computer program stored on a computer readable storage medium, the computer program including instructions for causing a processor of a master node to perform the following steps, wherein the steps are:
Receiving a distributed resource request signal for model learning from a client;
determining a node to perform learning of the model among the master node and one or more subnodes included in one cluster, upon receiving the distributed resource request signal;
obtaining learning result information according to learning of the model, and updating the model based on the learning result information; and
completing learning of the model and transmitting the updated model to the client when the updated model satisfies a preset condition;
Including,
When the updated model satisfies a preset condition, completing the learning of the model and transmitting the updated model to the client,
Transmitting optimal learning setting information together with the updated model;
including,
A computer program stored on a computer readable storage medium.
A computer program stored on a computer readable storage medium, the computer program including instructions for causing a processor of a master node to perform the following steps, wherein the steps are:
Receiving a distributed resource request signal for model learning from a client;
determining a node to perform learning of the model among the master node and one or more subnodes included in one cluster, upon receiving the distributed resource request signal;
obtaining learning result information according to learning of the model, and updating the model based on the learning result information; and
completing learning of the model and transmitting the updated model to the client when the updated model satisfies a preset condition;
Including,
The distributed resource request signal for learning the model,
At least one of identification information capable of identifying the client, learning setting information required for learning of the model, learning target information that is a learning goal of the model, and information about a job performed using the model to be learned,
A computer program stored on a computer readable storage medium.
15. The method of claim 14,
The learning setting information,
Including hyperparameter information related to the model,
A computer program stored on a computer readable storage medium.
A computer program stored on a computer readable storage medium, the computer program including instructions for causing a processor of a master node to perform the following steps, wherein the steps are:
Receiving a distributed resource request signal for model learning from a client;
determining a node to perform learning of the model among the master node and one or more subnodes included in one cluster, upon receiving the distributed resource request signal;
obtaining learning result information according to learning of the model, and updating the model based on the learning result information; and
completing learning of the model and transmitting the updated model to the client when the updated model satisfies a preset condition;
Including,
The learning result information,
Including at least one of version information of the learned model, weight of the learned model, weight change amount according to the learning, and information on a learning method of the model,
A computer program stored on a computer readable storage medium.
17. The method of claim 16,
The weight of the learned model is,
Reflecting knowledge extracted from a training data set used to train the model at the master node and the one or more subnodes,
A computer program stored on a computer-readable storage medium.

Images