KR102662645B1 - Method for active learning based on curriculum - Google Patents

Abstract

According to an embodiment of the present disclosure, a curriculum-based active learning method performed by a computing device is disclosed. The method includes training a neural network model based on a first learning data set among training data sets acquired through active learning; And it may include training the neural network model using a second learning data set that has a higher learning difficulty than the first learning data set among the learning data sets acquired through the active learning.

Description

Curriculum-based active learning method {METHOD FOR ACTIVE LEARNING BASED ON CURRICULUM}

The present invention relates to a learning method for a neural network model, and more specifically, to an improved active learning method for a neural network model.

In order to ensure good performance of deep learning, it is essential to secure a lot of high-quality learning data so that the neural network model can be sufficiently learned and operated. Therefore, in order to ensure good performance of deep learning, a lot of labeling costs are inevitably incurred during the learning process. One of the methods to improve this labeling cost problem is active learning.

Active learning refers to a learning method that uses a learned neural network model to select data effective for learning from an unlabeled data set and utilize it for learning. Specifically, in active learning, when a model trained based on a labeled data set selects data important for learning, an oracle labels the selected data and learns it using a model trained on an unlabeled data set. It is carried out as a process that reflects on . In other words, active learning consists of a training phase, which trains a neural network model using an existing labeled data set, and an acquisition phase, which is a process of extracting data to be labeled from a data set for which no label exists. can be distinguished.

Meanwhile, active learning is an effective learning strategy that can reduce the cost of labeling data that requires training of a neural network model, but it has several disadvantages because it generally uses a small amount of data sets in the learning stage of the neural network model. For example, general active learning uses a small data set in the learning stage, so the model is greatly affected by the initialization of the neural network and has the disadvantage of not being able to perform properly. In addition, general active learning uses a small amount of data sets in the learning stage, so there is a disadvantage in that the model is overfitted and cannot achieve proper performance.

Republic of Korea Patent Publication No. 10-2021-0075843 (2021.06.23) discloses a method for learning a neural network model.

The present disclosure was developed in response to the above-described background technology, and its purpose is to provide a learning method that can effectively improve the shortcomings of general active learning.

According to an embodiment of the present disclosure for realizing the above-described task, a curriculum-based active learning method performed by a computing device is disclosed. The method includes training a neural network model based on a first learning data set among training data sets acquired through active learning; And it may include training the neural network model using a second learning data set that has a higher learning difficulty than the first learning data set among the learning data sets acquired through the active learning.

In an alternative embodiment, the learning difficulty of the learning data set acquired through the active learning may be differentiated depending on the time of acquisition of the data through the active learning.

In an alternative embodiment, the second training data set may include a data set acquired more recently than the first training data set in the process of active learning.

In an alternative embodiment, training a neural network model based on the first training data set includes performing knowledge distillation of the neural network model based on a previously trained neural network model through active learning. It can be included.

In an alternative embodiment, the step of training the neural network model using the second training data set includes using the second training data after a predetermined point in the process of learning the neural network model based on the first training data set. It may include training the neural network model by adding a set as input.

In an alternative embodiment, the neural network model may be trained based on a first loss function applied to learning based on the first training data set and a second loss function applied to learning using the second training data set. there is.

In an alternative embodiment, the first loss function includes: a first sub-loss function for active learning of the neural network model based on the first training data set; and a second sub-loss function for distilling knowledge about the neural network model, which is performed based on a neural network model previously learned through active learning.

In an alternative embodiment, a first hyper parameter for controlling the knowledge distillation may be applied to the second sub-loss function.

In an alternative embodiment, the first hyperparameter may gradually decrease as training of the neural network model based on the first training data set progresses.

In an alternative embodiment, the first sub-loss function of the first loss function is a cross entropy function, and the second sub-loss function of the first loss function is the Kullback-Leibler divergence. divergence) function.

In an alternative embodiment, a second hyper-parameter may be applied to the second loss function to adjust the influence of the second training data set on learning of the neural network model.

In an alternative embodiment, the second hyperparameter may have a value of 0 at the initial time when training of the neural network model based on the first training data set is performed.

In an alternative embodiment, the second hyperparameter is gradually adjusted after a certain point in time to increase the influence of the second training data set in the process of learning the neural network model based on the first training data set. can be increased.

According to an embodiment of the present disclosure for realizing the above-described object, a computer program stored in a computer-readable storage medium is disclosed. When the computer program is executed on one or more processors, it performs the following operations to perform curriculum-based active learning, the operations being: based on a first learning data set among the learning data sets acquired through active learning. The operation of learning a neural network model; And it may include an operation of training the neural network model based on a second learning data set that has a higher learning difficulty than the first learning data set among the learning data sets acquired through the active learning.

According to an embodiment of the present disclosure for realizing the above-described task, a computing device that performs curriculum-based active learning based on deep learning is disclosed. The device includes a processor including at least one core; a memory containing program codes executable on the processor; and a network unit, wherein the processor trains a neural network model based on a first learning data set among the learning data sets acquired through active learning, and uses the first learning data among the learning data sets acquired through active learning. The neural network model can be trained based on a second learning data set that has a higher learning difficulty compared to the set.

In order to realize the above-described task, a computer-readable recording medium storing a data structure corresponding to the parameters of a neural network, at least partially updated during a learning process, is disclosed. The calculation of the neural network is at least partially based on the parameters, and the learning process includes: training a neural network model based on a first learning data set among training data sets acquired through active learning; And it may include training the neural network model based on a second learning data set that has a higher learning difficulty than the first learning data set among the learning data sets acquired through the active learning.

The present disclosure can provide a learning method that can effectively improve the shortcomings of general active learning.

1 is a block diagram of a computing device for performing curriculum-based active learning according to an embodiment of the present disclosure.
Figure 2 is a schematic diagram showing a neural network according to an embodiment of the present disclosure.
Figure 3 is a block diagram showing a typical existing active learning process.
Figure 4 is a block diagram showing an active learning process according to an embodiment of the present disclosure.
Figure 5 is a flowchart showing a learning process of a neural network model performed by a computing device according to an embodiment of the present disclosure.
Figure 6 is a schematic diagram of a computing environment according to one embodiment of the present disclosure.

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

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

Additionally, the term “or” is intended to mean an inclusive “or” and not an exclusive “or.” That is, unless otherwise specified or clear from context, “X utilizes A or B” is intended to mean one of the natural implicit substitutions. That is, either X uses A; X uses B; Or, if X uses both A and B, “X uses A or B” can apply to either of these cases. Additionally, the term “and/or” as used herein should be understood to refer to and include all possible combinations of one or more of the related listed items.

Additionally, the terms “comprise” and/or “comprising” should be understood to mean that the corresponding feature and/or element is present. However, the terms “comprise” and/or “comprising” should be understood as not excluding the presence or addition of one or more other features, elements and/or groups thereof. Additionally, unless otherwise specified or the context is clear to indicate a singular form, the singular terms herein and in the claims should generally be construed to mean “one or more.”

And, the term “at least one of A or B” should be interpreted to mean “a case containing only A,” “a case containing only B,” and “a case of combining A and B.”

Meanwhile, the term "Nth (N is a natural number)" expressed throughout the detailed description and claims of the present disclosure can be understood as an expression used to distinguish the components of the invention from a temporal, functional or structural perspective. . The term “Nth” can be understood as a term used to distinguish and express components according to a predetermined standard. For example, in the present disclosure, learning data sets classified according to the learning difficulty of a neural network model may be distinguished through the expression N-th configuration, such as a first learning data set, a second learning data set, etc. In addition, the loss function used for learning a neural network model can be distinguished according to the type through the expression "Nth configuration," such as a first loss function, a second loss function, etc. However, this is only one example, and the term “Nth” may be used to distinguish components according to various criteria from various perspectives.

Those skilled in the art will additionally recognize that the various illustrative logical blocks, components, modules, circuits, means, logic, and algorithm steps described in connection with the embodiments disclosed herein may be implemented using electronic hardware, computer software, or a combination of both. It must be recognized that it can be implemented with 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 software will depend on the specific application and design constraints imposed on the overall system. A skilled technician can implement the described functionality in a variety of ways for each specific application. However, such implementation decisions should not be construed as causing a departure from the scope of the present disclosure.

The description of the presented embodiments is provided to enable anyone skilled in the art to use or practice the present invention. Various modifications to these embodiments will be apparent to those skilled in the art. The general principles defined herein may be applied to other embodiments without departing from the scope of the disclosure. Therefore, the present invention is not limited to the embodiments presented herein. The present invention is to be interpreted in the broadest scope consistent with the principles and novel features presented herein.

1 is a block diagram of a computing device for performing curriculum-based active learning according to an embodiment of the present disclosure.

The configuration of the computing device 100 shown in FIG. 1 is only a simplified example. In one embodiment of the present disclosure, the computing device 100 may include different components for performing the computing environment of the computing device 100, and only some of the disclosed components may configure the computing device 100.

The computing device 100 may include a processor 110, a memory 130, and a network unit 150.

The processor 110 may be composed of one or more cores, and may include a central processing unit (CPU), a general purpose graphics processing unit (GPGPU), and a tensor processing unit (TPU) of a computing device. unit) may include a processor for data analysis and deep learning. The processor 110 may read a computer program stored in the memory 130 and perform data processing for machine learning according to an embodiment of the present disclosure. According to an embodiment of the present disclosure, the processor 110 may perform an operation for learning a neural network. The processor 110 can perform calculations for learning a neural network, such as processing input data for learning in deep learning, extracting features from input data, calculating errors, and updating the weights of the neural network using backpropagation. . At least one of the CPU, GPGPU, and TPU of the processor 110 may process learning of the neural network. For example, the CPU and GPGPU can work together to process neural network learning and data classification using neural networks. Additionally, in one embodiment of the present disclosure, the processors of a plurality of computing devices can be used together to process learning of a neural network and data classification using a neural network. Additionally, a computer program executed in a computing device according to an embodiment of the present disclosure may be a CPU, GPGPU, or TPU executable program.

According to one embodiment of the present disclosure, the processor 110 may learn a neural network model based on improved active learning. Here, improved active learning can be understood as a learning method applied to the learning stage in general active learning, which is divided into a learning stage and an acquisition stage. Specifically, the processor 110 may learn a neural network model based on a curriculum that gradually increases the difficulty of learning in the learning phase of active learning.

For example, the processor 110 performs an active learning step so that the neural network model sequentially learns a data set with a low learning difficulty (i.e. a data set that is easy to learn) to a data set with a high learning difficulty (i.e. a data set that is difficult to learn). You can control it. At this time, the difficulty of the learning data set can be classified according to the acquisition time of the data acquired through the acquisition stage of active learning. Since active learning expands the data set required for learning by repeatedly performing the learning and acquisition steps, the most recently acquired data (i.e. data labeled through the acquisition step immediately before learning) is used to create new data that the neural network model has not learned. As learning data, it can be understood as data with the highest learning difficulty. Accordingly, the processor 110 may classify the learning data set into a data set that is easier to learn as the data set is acquired older (i.e., a data set that is acquired more recently is more difficult to learn). Additionally, the processor 110 may control the learning process of the neural network model by inputting the classified learning data set into the neural network model according to the curriculum so that the neural network model performs learning while gradually increasing the learning difficulty. Through this curriculum-based active learning process, the impact of initialization of the deep neural network can be minimized, and a neural network model with the advantages of generalization and fast convergence speed can be built.

In addition, in the process of performing the above-described curriculum-based active learning, the processor 110 performs knowledge distillation of the neural network model to be learned in the current learning step based on the neural network model previously learned in the previous learning step of active learning. ) can be performed. Here, knowledge distillation can be understood as a computational process performed based on the timing of the learning phase in active learning in which the learning and acquisition phases are repeatedly performed, rather than the size of the neural network model.

For example, the processor 110 may transfer the knowledge of the neural network model previously learned in the previous learning step of active learning to the neural network model to be learned in the current learning step based on the neural network model previously learned in the previous learning step of active learning. there is. Based on the time of the learning phase of active learning, the processor 110 considers the model in the current learning phase as a student model, the model in the previous phase of the current learning phase as a teacher model, and considers the model in the current learning phase as a teacher model. By comparing the output with the output of the student model, knowledge distillation can be performed to train the student model to imitate the teacher model. This kind of knowledge distillation allows learning on data sets with low learning difficulty to be performed effectively during the learning stage of active learning. In the pre-learning stage of active learning, the neural network model learns on a data set with a low learning difficulty that has been acquired for a long time. Therefore, knowledge distillation using a neural network model that has been trained in the previous learning stage of active learning helps the neural network model to be learned in the current learning stage of active learning to effectively perform learning on a data set with low learning difficulty. Additionally, knowledge distillation using a neural network model learned in the prior learning stage of active learning helps prevent the neural network model from overfitting to data sets with low learning difficulty.

According to an embodiment of the present disclosure, the memory 130 may store any type of information generated or determined by the processor 110 and any type of information received by the network unit 150.

According to an embodiment of the present disclosure, the memory 130 is a flash memory type, hard disk type, multimedia card micro type, or card type memory (e.g. (e.g. SD or -Only Memory), and may include at least one type of storage medium among magnetic memory, magnetic disk, and optical disk. The computing device 100 may operate in connection with web storage that performs a storage function of the memory 130 on the Internet. The description of the memory described above is merely an example, and the present disclosure is not limited thereto.

The network unit 150 according to an embodiment of the present disclosure may use any type of known wired or wireless communication system.

The network unit 150 can transmit and receive data sets necessary for learning or inferring a neural network model, information processed by the processor 110, and a user interface through communication with other terminals or systems. For example, the network unit 150 may receive a data set necessary for learning a neural network model through communication with an external database. The network unit 150 may transmit the output of the neural network model learned by the processor 110 to an external database.

Meanwhile, the computing device 100 according to an embodiment of the present disclosure is a computing system that transmits and receives data through communication with a client and may include a server. At this time, the client may be any type of terminal that can access the server. For example, the computing device 100, which is a server, may train a plurality of neural network models through a training data set. The computing device 100, which is a server, may provide output data to the client using learned models according to the client's request. Additionally, the computing device 100, which is a server, may provide learned models to the client according to the client's request.

In an additional embodiment, the computing device 100 may include any type of terminal that receives data resources generated by an arbitrary server and performs additional information processing.

Figure 2 is a schematic diagram showing a neural network according to an embodiment of the present disclosure.

A neural network model according to an embodiment of the present disclosure may include a neural network for performing specific tasks such as detection, classification, segmentation, etc. A neural network can generally consist of a set of interconnected computational units, which can be referred to as nodes. These nodes may also be referred to as neurons. A neural network consists of at least one node. Nodes (or neurons) that make up neural networks may be interconnected by one or more links.

Within a neural network, one or more nodes connected through a link may form a relative input node and output node relationship. The concepts of input node and output node are relative, and any node in an output node relationship with one node may be in an input node relationship with another node, and vice versa. As described above, input node to output node relationships can be created around links. One or more output nodes 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 data of the output node may be determined based on the data input to the input node. Here, the link connecting the input node and the output node may have a weight. Weights may be variable and may be varied by the user or algorithm in order for the neural network to perform the desired function. For example, when one or more input nodes are connected to one output node by respective links, the output node is set to the values input to the input nodes connected to the output node and the links corresponding to each input node. The output node value can be determined based on the weight.

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

A neural network may consist of a set of one or more nodes. A subset of nodes that make up a neural network can form a layer. Some of the nodes constituting the neural network may form one layer based on the distances from the first input node. For example, a set of nodes with a distance n from the initial input node may constitute n layers. The distance from the initial input node can be defined by the minimum number of links that must be passed to reach the node from the initial input node. However, this definition of a layer is arbitrary for explanation purposes, and the order of a layer within a neural network may be defined in a different way than described above. For example, a layer of nodes may be defined by distance from the final output node.

The initial input node may refer to one or more nodes in the neural network through which data is directly input without going through links in relationships with other nodes. Alternatively, in the relationship between nodes based on links within a neural network, it may refer to nodes that do not have other input nodes connected by links. Similarly, the final output node may refer to one or more nodes that do not have an output node in their relationship with other nodes among the nodes in the neural network. Additionally, hidden nodes may refer to nodes constituting a neural network other than the first input node and the last output node.

The neural network according to an embodiment of the present disclosure is a neural network in which 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 it progresses from the input layer to the hidden layer. You can. In addition, the neural network according to another embodiment of the present disclosure may be a neural network in which the number of nodes in the input layer may be less than the number of nodes in the output layer, and the number of nodes decreases as it progresses from the input layer to the hidden layer. there is. In addition, the neural network according to another embodiment of the present disclosure may be 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 it progresses from the input layer to the hidden layer. You can. A neural network according to another embodiment of the present disclosure may be a neural network that is a combination of the above-described neural networks.

A deep neural network (DNN) may refer to a neural network that includes multiple hidden layers in addition to the input layer and output layer. Deep neural networks allow you to identify latent structures in data. In other words, it is possible to identify the potential structure of a photo, text, video, voice, or music (e.g., what object is in the photo, what the content and emotion of the text are, what the content and emotion of the voice are, etc.) . Deep neural networks include convolutional neural networks (CNN), recurrent neural networks (RNN), auto encoders, generative adversarial networks (GAN), and restricted Boltzmann machines (RBM). machine), deep belief network (DBN), Q network, U network, Siamese network, etc. The description of the deep neural network described above is only an example and the present disclosure is not limited thereto.

In one embodiment of the present disclosure, the neural network may include an autoencoder. An autoencoder may be a type of artificial neural network to output output data similar to input data. The autoencoder may include at least one hidden layer, and an odd number of hidden layers may be placed between input and output layers. The number of nodes in each layer may be reduced from the number of nodes in the input layer to an intermediate layer called the bottleneck layer (encoding), and then expanded symmetrically and reduced from the bottleneck layer to the output layer (symmetrical to the input layer). Autoencoders can perform nonlinear dimensionality reduction. The number of input layers and output layers can be corresponded to the dimension after preprocessing of the input data. In an auto-encoder structure, the number of nodes in the hidden layer included in the encoder may have a structure that decreases as the distance from the input layer increases. The number of nodes in the bottleneck layer (the layer with the fewest nodes located between the encoder and decoder) may not be able to convey a sufficient amount of information if it is too small, so it must exceed a certain number (e.g., more than half of the input layers, etc.). ) may be maintained.

A neural network may be trained in at least one of supervised learning, unsupervised learning, semi-supervised learning, or reinforcement learning. Learning of a neural network may be a process of applying knowledge for the neural network to perform a specific operation to the neural network.

Neural networks can be trained to minimize output errors. In neural network learning, learning data is repeatedly input into the neural network, the output of the neural network and the error of the target for the learning data 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. This is the process of updating the weight of each node in the neural network through backpropagation. In the case of teacher learning, learning data in which the correct answer is labeled in each learning data is used (i.e., labeled learning data), and in the case of non-teacher learning, the correct answer may not be labeled in each learning data. That is, for example, in the case of teacher learning regarding data classification, the learning data may be data in which each learning data is labeled with a category. Labeled training data is input to the neural network, and the error can be calculated by comparing the output (category) of the neural network with the label of the training data. As another example, in the case of non-teachable learning for data classification, the error can be calculated by comparing the input training data with the neural network output. The calculated error is back-propagated in the neural network in the reverse direction (i.e., from the output layer to the input layer), and the connection weight of each node in each layer of the neural network can be updated according to back-propagation. The amount of change in the connection weight of each updated node may be determined according to the learning rate. The neural network's calculation of input data and backpropagation of errors can constitute a learning cycle (epoch). The learning rate may be applied differently depending on the number of repetitions of the learning cycle of the neural network. For example, in the early stages of learning a neural network, a high learning rate can be used to ensure that the neural network quickly achieves a certain level of performance to increase efficiency, and in the later stages of training, a low learning rate can be used to increase accuracy.

In the learning of neural networks, the training data can generally be a subset of real data (i.e., the data to be processed using the learned neural network), and thus the error for the training data is reduced, but the error for the real data is reduced. There may be an incremental learning cycle. Overfitting is a phenomenon in which errors in actual data increase due to over-learning from the training data. For example, a phenomenon in which a neural network that learned a cat by showing a yellow cat fails to recognize that it is a cat when it sees a non-yellow cat may be a type of overfitting. Overfitting can cause errors in machine learning algorithms to increase. To prevent such overfitting, various optimization methods can be used. To prevent overfitting, methods such as increasing the learning data, regularization, dropout that disables some of the network nodes during the learning process, and use of a batch normalization layer can be applied. there is.

Figure 3 is a block diagram showing a typical existing active learning process.

Active learning basically involves a training phase in which a neural network model is trained using an existing data set (i.e. labeled data set) and a training phase in which data to be assigned a label is selected from a data pool for which no label exists. The process can be distinguished into an acquisition phase. Existing general active learning has an iterative structure in which the data set obtained through the acquisition stage is added to the existing data set, and the learning stage is performed again. Specifically, in the learning phase of existing active learning, a neural network model can perform learning based on a labeled data set. At this time, the labeled data set may be a data set initially labeled through an oracle, or the data selected through the acquisition stage and labeled through an oracle may be a data set that is additionally included in the existing labeled data set. . Here, an oracle can be understood as any module that performs labeling or annotation based on queries from a learned neural network model. When the learning stage is completed, in the acquisition stage of existing active learning, the neural network model can generate a query to select data necessary for learning based on the learning results according to the query strategy. And, in the acquisition phase of existing active learning, the oracle can select data and assign labels according to the query from a data pool for which no label exists. The learning and acquisition steps of existing active learning as described above can be performed repeatedly until the neural network model reaches the target performance.

For example, referring to FIG. 3, in the initial learning step for existing active learning, the neural network model M (200) may receive a labeled learning data set L ₀ and perform learning. Neural network model M (200) can generate a query to select data required for next learning based on the initial learning result. When the initial learning step is completed, in the first learning step, Oracle selects the data required for the second learning from the data pool for which no label exists based on the query and performs labeling to create the acquired data set G ₀ . When the first learning step is completed, in the second learning step, the neural network model M (200) can construct a learning data set L ₁ by including the acquired data set G ₀ in the existing learning data set L ₀ and perform learning based on it. there is. And, the neural network model M (200) may generate a query to select data required for next learning based on the second learning result. When the second learning step is completed, in the second learning step, Oracle selects the data required for the third learning from the data pool for which no label exists based on the query and performs labeling to create the acquired data set G ₁ . When the second learning step is completed, in the third learning step, the neural network model M (200) can construct a learning data set by including the acquired data set G ₁ in the existing learning data set L ₁ and perform learning based on it. Such learning and acquisition can be performed repeatedly until the neural network model M (200) reaches the target performance.

Figure 4 is a block diagram showing an active learning process according to an embodiment of the present disclosure.

The existing active learning shown in Figure 3 suffers from chronic problems in that it is greatly affected by the initialization of the neural network and the neural network is prone to overfitting. Active learning according to an embodiment of the present disclosure, as shown in FIG. 4, can effectively improve this problem by utilizing curriculum and knowledge distillation.

In active learning according to an embodiment of the present disclosure, a neural network model may perform learning based on a curriculum that gradually increases the difficulty of the learning data set in the learning stage. At this time, an indicator that classifies the difficulty of the learning data set may be the acquisition time of the learning data set. Therefore, the curriculum can be understood as a concept of dividing the difficulty of the learning data set according to the acquisition time of the learning data set and training a neural network model by increasing the learning difficulty based on the divided learning data set. In active learning, recently acquired data (i.e. data most recently labeled based on learning results and included in the learning data set) is data that the neural network model has not used for training, so it is compared to the previously acquired and already trained data set. It can be considered the most difficult. Therefore, active learning according to an embodiment of the present disclosure is a curriculum that divides data sets previously acquired and used for learning into data that are easy to learn, and data sets that were recently acquired and not used for learning as data that are difficult to learn. It can be done based on In other words, the neural network model according to an embodiment of the present disclosure can perform curriculum-based learning that gradually increases the learning difficulty by sequentially receiving data sets from the oldest acquisition time to the recently acquired data set and performing learning. there is.

Additionally, in active learning according to an embodiment of the present disclosure, knowledge distillation may be performed in which the knowledge of the neural network model in the previous learning step is transferred to the neural network model in the current learning step. At this time, knowledge distillation can be understood as a technical concept that allows a neural network model that performs curriculum-based active learning to perform efficient learning by receiving knowledge already learned in the previous learning stage. As mentioned above, the difficulty of a learning data set is differentiated depending on the acquisition time of the learning step, so knowledge distillation helps to effectively perform learning on a data set with low learning difficulty. In other words, in curriculum-based active learning, the neural network model performs initial learning based on the data set learned in the previous learning step, which is a learning data set with low learning difficulty, so knowledge distillation allows the neural network model to use the knowledge learned in the previous learning step. Helps you adapt quickly.

For example, active learning according to an embodiment of the present disclosure has a basic structure divided into a learning stage and an acquisition stage as shown in FIG. 3, and the learning stage and the acquisition stage are repeatedly performed until the neural network model reaches the target performance. ensure that it is carried out. Referring to FIG. 4, in the first learning step for active learning, the neural network model M ₁ 310 may receive a labeled learning data set L ₀ and perform learning. When the initial acquisition step of generating the acquisition data set G ₀ by selecting the data required for the second learning from the data pool for which no label exists and performing labeling is completed, in the second learning step, the neural network model M ₂ (320) is used in the existing learning step. By including the acquired data set G ₀ in the data set L ₀ , a learning data set L ₁ can be formed and learning can be performed based on it. When the second acquisition step of generating the acquisition data set G ₁ by selecting the data required for the third learning from the data pool for which no label exists and performing labeling is completed, in the third learning step, the neural network model M ₃ 330 is used in the existing learning step. You can construct a learning data set by including the acquired data set G ₁ in the data set L ₁ and perform learning based on it. At this time, the neural network model M ₁ (310), the neural network model M ₂ (320), and the neural network model M ₃ (330) are one neural network model that performs active learning, differentiated according to the learning time to explain knowledge distillation. It can be understood. Since active learning according to an embodiment of the present disclosure is a technical concept applied to the learning stage, the detailed description of the acquisition stage will be replaced with the above-described content based on FIG. 3.

However, unlike the existing active learning of FIG. 3, active learning according to an embodiment of the present disclosure performs curriculum-based learning in the learning phase and distills previously learned knowledge in the learning phase. Knowledge distillation and curriculum can all be applied equally to the iterative learning steps in active learning. Below, we describe the knowledge distillation and curriculum-based learning process based on the third learning stage. Specifically, at the beginning (10) of the learning stage, a neural network is run based only on the data set L ₁ used in the existing learning stage as a data set with low learning difficulty, excluding the most recently acquired data set G _1, which is a data set with high learning difficulty. Learning of model M ₃ (330) may proceed. At this time, in order to enable the neural network model M ₃ (330) to effectively learn the data set L ₁ that is easy to learn, the neural network model M ₂ (320) learned in the previous second learning step is used to train the neural network being trained in the current learning step. Knowledge distillation can be performed on Model M ₃ (330). This knowledge distillation serves to prevent the neural network model M ₃ 330 from overfitting to a data set that is easy to learn. When learning based on the data set L ₁ with low learning difficulty reaches a certain level, _a neural _network model M ₃ ( 330) learning may take place in the latter half (20). At this time, whether learning has reached a certain level can be determined based on whether the number of cycle repetitions of learning based on the data set L ₁ with low learning difficulty is greater than or equal to a threshold. Whether learning has reached a certain level can be determined based on whether indicators for learning results, such as accuracy and reliability, are above a threshold. However, since the description of the above-described learning level judgment is only an example, various judgment methods can be applied within the range of changes that can be made by those skilled in the art based on the present disclosure.

Meanwhile, the above-described knowledge distillation and curriculum-based active learning process can be implemented by calculating the loss function as shown in [Equation 1] below.

[Equation 1]

Here, g _t-1 is a recently acquired data set (ie a data set with high learning difficulty), and L _t-1 is a data set used in the previous learning step (ie a data set with low learning difficulty). Additionally, x _i and y _i are training data and labels corresponding to the training data, respectively. L _ce is the cross entropy function, and L _dist is the Kullback-Leibler divergence function used for knowledge distillation. α is a hyperparameter to control knowledge distillation, and β is a hyperparameter to control the influence of the difficult-to-learn data set g _t-1 . At the beginning of the learning phase of active learning according to an embodiment of the present disclosure, the data set g _t-1 , which is difficult to learn, is not used for learning, so β may be set to a value of 0.

When the neural network model has been learned to a certain level based on the easy-to-learn data set L _t-1 , the most recently acquired data set g _t-1 through the curriculum is added to the learning data set to enable learning of the neural network model. It can proceed. This process can be implemented by gradually increasing the value of β. In other words, when the learning of the neural network model reaches a certain level during the process of learning based on the data set L _t-1 with low learning difficulty, the neural network model can perform learning on the data set g _t-1 with high learning difficulty. To enable this, the value of β can be increased sequentially.

In order to induce the neural network model to sufficiently learn the data set g _t-1 with high learning difficulty, it is necessary to reduce the learning influence of the loss function L _dist corresponding to knowledge distillation as learning progresses. Therefore, as learning based on the easy-to-learn data set L _t-1 progresses, the proportion of the loss function L _dist corresponding to knowledge distillation can be gradually reduced in the overall loss function. This process can be implemented by gradually decreasing the value of α in [Equation 1]. As the value of α gradually decreases, 1-α increases on the contrary, so as the proportion of the loss function L _dist decreases, learning based on data L _t-1 with low learning difficulty expressed as the sum of the loss function L _dist is used. The proportion of the loss function used (i.e. L _ce multiplied by 1-α) can be increased.

Figure 5 is a flowchart showing a learning process of a neural network model performed by a computing device according to an embodiment of the present disclosure.

Referring to FIG. 5 , in step S100, the computing device 100 according to an embodiment of the present disclosure may learn a neural network model based on a first learning data set among the training data sets acquired through active learning. Learning of a neural network model based on the first training data set may be performed in the learning stage of active learning or the learning stage of the acquisition stage. The first learning data set may be a data set with low learning difficulty among learning data acquired through active learning. A low learning difficulty can be understood as a long period of data acquisition through active learning.

In step S100, the computing device 100 may perform knowledge distillation of the neural network model based on a neural network model previously learned through active learning. For example, assuming that the learning time point in step S100 is the first time point and the learning time point in the learning step performed before step S100 is the second time point, the computing device 100 may use the neural network model learned at the second time point. Knowledge distillation of the neural network model can be performed by transferring knowledge to the neural network model at the first time point. At this time, the neural network model learned at the second viewpoint and the neural network model at the first viewpoint may be the same neural network model that performs active learning, with only a difference in viewpoint. That is, the computing device 100 may perform knowledge distillation for learning of the first learning data set in step S100 so that the neural network model can utilize the knowledge learned in the past learning step in the current learning step. Knowledge distillation may be performed prior to the learning process for the first learning data set described above, or may be performed in parallel.

In step S200, the computing device 100 may train a neural network model using a second learning data set that has a higher learning difficulty than the first learning data set among the training data sets acquired through active learning. Learning of a neural network model based on the second learning data set may be performed in the learning stage of active learning or the learning stage of the acquisition stage. The second learning data set may be a data set with a high learning difficulty among learning data acquired through active learning. In other words, the second learning data set can be understood as including a data set acquired more recently than the first learning data set in the process of active learning.

Specifically, in step S200, the computing device 100 may learn a neural network model by gradually adding the second learning data set as input after a predetermined point in the process of step S100. If the level of learning based on the first learning data set is higher than a predefined reference value, in step S200, the computing device 100 uses the first learning data set as well as the second learning data set as input to the neural network model to create a neural network model. can be learned. Accordingly, the predetermined point in time can be understood as the point in time when the level of learning based on the first learning data set reaches the reference value.

Meanwhile, in steps S100 and S200, the neural network model may be learned based on a first loss function applied to learning based on the first training data set and a second loss function applied to learning using the second training data set. . At this time, the first loss function is a first sub-loss function for learning the neural network model based on the first learning data set and a second sub-loss function for knowledge distillation based on a neural network model previously learned through active learning. may include. The first loss function may be defined as the weighted sum of the first sub-loss function and the second sub-loss function. For a weighted sum of the first sub-loss function and the second sub-loss function, the first hyper-parameter may be applied to the second sub-loss function. Additionally, a second hyperparameter, which is a coefficient to independently control the proportion within the overall loss function, may be applied to the second loss function.

For example, referring to [Equation 1], the first sub-loss function and the second loss function may be the cross-entropy function L _ce , and the second sub-loss function may be the Kullback-Leibler divergence function L _dist . The first hyperparameter is a coefficient for controlling knowledge distillation and may be α applied to the second sub-loss function. The first hyperparameter may gradually decrease as learning of the neural network model based on the first training data set progresses. The second hyperparameter is a coefficient for controlling the influence of the second learning data set on learning of the neural network model, and may be β applied to the second sub-loss function. The second hyperparameter may have a value of 0 at the first time when learning of the neural network model based on the first training data set is performed. The second hyperparameter may be gradually increased after a certain point in time to increase the influence of the second learning data set in the process of learning the neural network model based on the first training data set. Here, the predetermined point in time can be understood as the point in time when the learning result of the neural network model based on the first learning data set reaches the target value. The target value may be the number of repetitions of the learning cycle, or it may be an indicator of learning results such as accuracy and reliability. Since the above-mentioned target value is only an example, the target value may be configured in various ways within a range that can be changed by a person skilled in the art based on the present disclosure.

Meanwhile, according to an embodiment of the present disclosure, a computer-readable medium storing a data structure is disclosed.

Data structure can refer to the organization, management, and storage of data to enable efficient access and modification of data. Data structure can refer to the organization of data to solve a specific problem (e.g., retrieving data, storing data, or modifying data in the shortest possible time). A data structure may be defined as a physical or logical relationship between data elements designed to support a specific data processing function. Logical relationships between data elements may include connection relationships between user-defined data elements. Physical relationships between data elements may include actual relationships between data elements that are physically stored in a computer-readable storage medium (e.g., a persistent storage device). A data structure may specifically include a set of data, relationships between data, and functions or instructions applicable to the data. Effectively designed data structures allow computing devices to perform computations while minimizing the use of the computing device's resources. Specifically, computing devices can increase the efficiency of operations, reading, insertion, deletion, comparison, exchange, and search through effectively designed data structures.

Data structures can be divided into linear data structures and non-linear data structures depending on the type of data structure. A linear data structure may be a structure in which only one piece of data is connected to another piece of data. Linear data structures may include List, Stack, Queue, and Deque. A list can refer to a set of data that has an internal order. The list may include a linked list. A linked list may be a data structure in which data is connected in such a way that each data is connected in a single line with a pointer. In a linked list, a pointer can contain connection information to the next or previous data. A linked list can be expressed as a singly linked list, doubly linked list, or circular linked list depending on its form. A stack may be a data listing structure that allows limited access to data. A stack can be a linear data structure in which data can be processed (for example, inserted or deleted) at only one end of the data structure. Data stored in the stack may have a data structure (LIFO-Last in First Out) where the later it enters, the sooner it comes out. A queue is a data listing structure that allows limited access to data. Unlike the stack, it can be a data structure (FIFO-First in First Out) where data stored later is released later. A deck can be a data structure that can process data at both ends of the data structure.

A non-linear data structure may be a structure in which multiple pieces of data are connected behind one piece of data. Nonlinear data structures may include graph data structures. A graph data structure can be defined by vertices and edges, and an edge can include a line connecting two different vertices. Graph data structure may include a tree data structure. A tree data structure may be a data structure in which there is only one path connecting two different vertices among a plurality of vertices included in the tree. In other words, it may be a data structure that does not form a loop in the graph data structure.

Data structures may include neural networks. And the data structure including the neural network may be stored in a computer-readable medium. Data structures including neural networks also include data preprocessed for processing by a neural network, data input to the neural network, weights of the neural network, hyperparameters of the neural network, data acquired from the neural network, activation functions associated with each node or layer of the neural network, neural network It may include a loss function for learning. A data structure containing a neural network may include any of the components disclosed above. In other words, the data structure including the neural network includes preprocessed data for processing by the neural network, data input to the neural network, weights of the neural network, hyperparameters of the neural network, data acquired from the neural network, activation functions associated with each node or layer of the neural network, neural network It may be composed of all or any combination of loss functions for learning. In addition to the configurations described above, a data structure containing a neural network may include any other information that determines the characteristics of the neural network. Additionally, the data structure may include all types of data used or generated in the computational process of a neural network and is not limited to the above. Computer-readable media may include computer-readable recording media and/or computer-readable transmission media. A neural network can generally consist of a set of interconnected computational units, which can be referred to as nodes. These nodes may also be referred to as neurons. A neural network consists of at least one node.

The data structure may include data input to the neural network. A data structure containing data input to a neural network may be stored in a computer-readable medium. Data input to the neural network may include learning data input during the neural network learning process and/or input data input to the neural network on which training has been completed. Data input to the neural network may include data that has undergone pre-processing and/or data subject to pre-processing. Preprocessing may include a data processing process to input data into a neural network. Therefore, the data structure may include data subject to preprocessing and data generated by preprocessing. The above-described data structure is only an example and the present disclosure is not limited thereto.

The data structure may include the weights of the neural network. (In this specification, weights and parameters may be used with the same meaning.) And the data structure including the weights of the neural network may be stored in a computer-readable medium. A neural network may include multiple weights. Weights may be variable and may be varied by the user or algorithm in order for the neural network to perform the desired function. For example, when one or more input nodes are connected to one output node by respective links, the output node is set to the values input to the input nodes connected to the output node and the links corresponding to each input node. Based on the weight, the data value output from the output node can be determined. The above-described data structure is only an example and the present disclosure is not limited thereto.

As an example and not a limitation, the weights may include weights that are changed during the neural network learning process and/or weights for which neural network learning has been completed. Weights that change during the neural network learning process may include weights that change at the start of the learning cycle and/or weights that change during the learning cycle. Weights for which neural network training has been completed may include weights for which a learning cycle has been completed. Therefore, the data structure including the weights of the neural network may include weights that are changed during the neural network learning process and/or the data structure including the weights for which neural network learning has been completed. Therefore, the above-mentioned weights and/or combinations of each weight are included in the data structure including the weights of the neural network. The above-described 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 (e.g., memory, 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 a different computing device and later reconstituted and used. Computing devices can transmit and receive data over a network by serializing data structures. Data structures containing the weights of a serialized neural network can be reconstructed on the same computing device or on a different 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 to increase computational efficiency while minimizing the use of computing device resources (e.g., in non-linear data structures, B-Tree, Trie, m-way search tree, AVL tree, Red-Black Tree) may be included. The foregoing is merely an example and the present disclosure is not limited thereto.

The data structure may include hyper-parameters of a neural network. And the data structure including the hyperparameters of the neural network can be stored in a computer-readable medium. A hyperparameter may be a variable that can be changed by the user. Hyperparameters include, for example, learning rate, cost function, number of learning cycle repetitions, weight initialization (e.g., setting the range of weight values subject to weight initialization), Hidden Unit. It may include a number (e.g., number of hidden layers, number of nodes in hidden layers). The above-described data structure is only an example and the present disclosure is not limited thereto.

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

Although the present disclosure has generally been described above as being capable of being implemented by a computing device, those skilled in the art will understand that the present disclosure can be implemented in combination with computer-executable instructions and/or other program modules that can be executed on one or more computers and/or in hardware and software. It will be well known that it can be implemented as a combination.

Typically, program modules include routines, programs, components, data structures, etc. that perform specific tasks or implement specific abstract data types. Additionally, those skilled in the art will understand that the methods of the present disclosure are applicable to single-processor or multiprocessor computer systems, minicomputers, mainframe computers, as well as personal computers, handheld computing devices, microprocessor-based or programmable consumer electronics, etc. It will be appreciated that each of these may be implemented in other computer system configurations, including those capable of operating in conjunction with one or more associated devices.

The described embodiments of the disclosure can 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.

Computers typically include a variety of computer-readable media. Computer-readable media can be any medium that can be accessed by a computer, and such computer-readable media includes volatile and non-volatile media, transitory and non-transitory media, removable and non-transitory media. Includes 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 refers to volatile and non-volatile media, transient and non-transitory media, removable and non-removable, 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, magnetic cassette, magnetic tape, magnetic disk storage or other magnetic storage. This includes, but is not limited to, a device, or any other medium that can be accessed by a computer and used to store desired information.

A computer-readable transmission medium typically implements computer-readable instructions, data structures, program modules, or other data on a modulated data signal, such as a carrier wave or other transport mechanism. Includes all information delivery media. The term modulated data signal refers to a signal in which one or more of the characteristics of the signal have been set or changed 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 example environment 1100 is shown that implements various aspects of the present disclosure, 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 processors and other multiprocessor architectures may also be used as processing unit 1104.

System bus 1108 may be any of several types of bus structures that may further be interconnected to a memory bus, peripheral bus, and 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. The basic input/output system (BIOS) is stored in non-volatile memory 1110, such as ROM, EPROM, and EEPROM, and is a basic input/output system that helps transfer information between components within the computer 1102, such as during startup. Contains routines. RAM 1112 may also include high-speed RAM, such as static RAM, for caching data.

Computer 1102 may also include an internal hard disk drive (HDD) 1114 (e.g., 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 a removable diskette 1118), and an optical disk drive 1120 (e.g., a CD-ROM for reading the disk 1122 or reading from or writing to other high-capacity optical media such as DVDs). Hard disk drive 1114, magnetic disk drive 1116, and optical disk drive 1120 are connected to system bus 1108 by hard disk drive interface 1124, magnetic disk drive interface 1126, and optical drive interface 1128, respectively. ) can be connected to. The interface 1124 for implementing an external drive includes 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. For computer 1102, drive and media correspond to storing any data in a suitable digital format. Although the description of computer-readable media above refers to removable optical media such as HDDs, removable magnetic disks, and CDs or DVDs, those skilled in the art will also recognize removable optical media such as zip drives, magnetic cassettes, flash memory cards, cartridges, etc. It will be appreciated that other types of computer-readable media, such as the like, may also be used in the example operating environment and that any such media may contain computer-executable instructions for performing the methods of the present disclosure.

A number of program modules may be stored in 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 on various commercially available operating systems or combinations of operating systems.

A user may enter commands and information into computer 1102 through one or more wired/wireless input devices, such as a keyboard 1138 and a pointing device such as mouse 1140. Other input devices (not shown) may include microphones, IR remote controls, joysticks, game pads, stylus pens, touch screens, etc. These and other input devices are connected to the processing unit 1104 through an input device interface 1142, which is often connected to the system bus 1108, but may also include a parallel port, an IEEE 1394 serial port, a game port, a USB port, an IR interface, It can be connected by other interfaces, etc.

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

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, computing device computer, router, personal computer, portable computer, microprocessor-based entertainment device, peer device, or other conventional network node, and is generally connected to computer 1102. For simplicity, only memory storage device 1150 is shown, although it includes many or all of the components described. The logical connections depicted include wired/wireless connections to a local area network (LAN) 1152 and/or a larger network, such as a wide area network (WAN) 1154. These LAN and WAN networking environments are common in offices and companies and facilitate enterprise-wide computer networks, such as intranets, all of which can be connected to a worldwide computer network, such as the Internet.

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

Computer 1102 may be associated with any wireless device or object deployed and operating in wireless communications, such as a printer, scanner, desktop and/or portable computer, portable data assistant (PDA), communications satellite, wirelessly detectable tag. Performs actions to communicate with any device or location and telephone. This includes at least Wi-Fi and Bluetooth wireless technologies. Accordingly, communication may be a predefined structure as in a conventional network or may simply be ad hoc communication between at least two devices.

Wi-Fi (Wireless Fidelity) allows connection to the Internet, etc. without wires. Wi-Fi is a wireless technology, like cell phones, that allows these devices, such as computers, to send and receive data indoors and outdoors, anywhere within the coverage area of a cell tower. Wi-Fi networks use wireless 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 wireless bands, for example, at data rates of 11 Mbps (802.11a) or 54 Mbps (802.11b), or in products that include both bands (dual band). .

Those skilled in the art will understand that information and signals may be represented using any of a variety of different technologies and techniques. For example, data, instructions, commands, information, signals, bits, symbols and chips that may be referenced in the above description include voltages, currents, electromagnetic waves, magnetic fields or particles, optical fields. It can be expressed by particles or particles, or any combination thereof.

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 may be used in electronic hardware, (for convenience) It will be understood that it 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 interoperability of hardware and software, various illustrative components, blocks, modules, circuits and steps have been described above generally with respect to their functionality. Whether this functionality is implemented as hardware or software depends on the specific application and design constraints imposed on the overall system. A person skilled in the art of this disclosure may implement the described functionality in various ways for each specific application, but such implementation decisions should not be construed as departing from the scope of this disclosure.

The 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, carrier, or media accessible from any computer-readable storage device. For example, computer-readable storage media include magnetic storage devices (e.g., hard disks, floppy disks, magnetic strips, etc.), optical disks (e.g., CDs, DVDs, etc.), smart cards, and flash. Includes, but is not limited to, memory devices (e.g., EEPROM, cards, sticks, key drives, etc.). Additionally, various storage media presented herein include one or more devices and/or other machine-readable media for storing information.

It is to be understood that the specific order or hierarchy of steps in the processes presented is an example of illustrative approaches. It is to be understood that the specific order or hierarchy of steps in processes may be rearranged within the scope of the present disclosure, based on design priorities. The appended method claims present elements of the various steps in a sample order but are not meant to be limited to the particular order or hierarchy presented.

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, and the general principles defined herein may be applied to other embodiments without departing from the scope of the disclosure. Thus, the present disclosure is not limited to the embodiments presented herein but is to be interpreted in the broadest scope consistent with the principles and novel features presented herein.

Claims (16)

A curriculum-based active learning method performed by a computing device including at least one processor, comprising:
Learning a neural network model based on a first learning data set among the learning data sets acquired through active learning; and
training the neural network model using a second learning data set having a higher learning difficulty than the first learning data set among the learning data sets acquired through the active learning;
Including,
The learning difficulty of the learning data set acquired through active learning is,
Distinguished depending on the time of acquisition of data through active learning,
The second learning data set has a higher learning difficulty compared to the first training data set,
Including a data set recently acquired compared to the first learning data set in the process of active learning,
The learning difficulty level differentiated according to the acquisition period is,
The neural network model is used in a curriculum-based learning process that performs learning while gradually increasing the learning difficulty,
The neural network model is,
Learned based on a first loss function applied to learning based on the first learning data set and a second loss function applied to learning using the second learning data set,
The first loss function is,
a first sub-loss function for learning the neural network model based on the first training data set; and
a second sub-loss function for distilling knowledge about the neural network model, which is performed based on a neural network model previously learned through active learning;
Including,
method.
delete delete According to claim 1,
The step of training a neural network model based on the first learning data set is:
performing knowledge distillation on the neural network model based on a neural network model previously learned through the active learning;
Including,
method.
According to claim 1,
The step of training the neural network model using the second training data set is:
training the neural network model by adding the second learning data set as an input after a predetermined point in the process of learning the neural network model based on the first training data set;
Including,
method.
delete delete According to claim 1,
A first hyper parameter for controlling the knowledge distillation is applied to the second sub-loss function,
method.
According to claim 8,
The first hyperparameter is,
Gradually decreases as learning of the neural network model based on the first learning data set progresses,
method.
According to claim 1,
The first sub-loss function of the first loss function is,
It is a cross entropy function,
The second sub-loss function of the first loss function is,
The Kullback-Leibler divergence function,
method.
According to claim 1,
A second hyperparameter is applied to the second loss function to adjust the influence of the second learning data set on learning of the neural network model.
method.
According to claim 11,
The second hyperparameter is,
Having a value of 0 at the first time when learning of the neural network model based on the first learning data set is performed,
method.
According to claim 11,
The second hyperparameter is,
In order to increase the influence of the second learning data set in the process of learning the neural network model based on the first learning data set, it is gradually increased after a certain point in time.
method.
A computer program stored in a computer-readable storage medium, wherein the computer program, when executed on one or more processors, performs the following operations for performing curriculum-based active learning, the operations being: :
An operation of training a neural network model based on a first learning data set among the learning data sets acquired through active learning; and
An operation of training the neural network model based on a second learning data set that has a higher learning difficulty than the first learning data set among the learning data sets acquired through the active learning;
Including,
The learning difficulty of the learning data set acquired through active learning is,
Distinguished depending on the time of acquisition of data through active learning,
The second learning data set has a higher learning difficulty compared to the first training data set,
Including a data set recently acquired compared to the first learning data set in the process of active learning,
The learning difficulty level differentiated according to the acquisition period is,
The neural network model is used in a curriculum-based learning process that performs learning while gradually increasing the learning difficulty,
The neural network model is,
Learned based on a first loss function applied to learning based on the first learning data set and a second loss function applied to learning using the second learning data set,
The first loss function is,
a first sub-loss function for learning the neural network model based on the first training data set; and
a second sub-loss function for distilling knowledge about the neural network model, which is performed based on a neural network model previously learned through active learning;
Including,
A computer program stored on a computer-readable storage medium.
A computing device that performs curriculum-based active learning,
A processor including at least one core;
a memory containing program codes executable on the processor; and
network department;
Including,
The processor,
Train a neural network model based on the first learning data set among the learning data sets acquired through active learning,
Training the neural network model based on a second learning data set that has a higher learning difficulty than the first learning data set among the learning data sets acquired through active learning,
The learning difficulty of the learning data set acquired through active learning is,
Distinguished depending on the time of acquisition of data through active learning,
The second learning data set has a higher learning difficulty compared to the first training data set,
Including a data set recently acquired compared to the first learning data set in the process of active learning,
The learning difficulty level differentiated according to the acquisition period is,
The neural network model is used in a curriculum-based learning process that performs learning while gradually increasing the learning difficulty,
The neural network model is,
Learned based on a first loss function applied to learning based on the first learning data set and a second loss function applied to learning using the second learning data set,
The first loss function is,
a first sub-loss function for learning the neural network model based on the first training data set; and
a second sub-loss function for distilling knowledge about the neural network model, which is performed based on a neural network model previously learned through active learning;
Including,
Device.
A computer-readable recording medium storing instructions,
The command causes the computing device to perform the following steps to perform curriculum-based active learning, which steps include:
Learning a neural network model based on a first learning data set among the learning data sets acquired through active learning; and
training the neural network model based on a second learning data set that has a higher learning difficulty than the first learning data set among the learning data sets acquired through the active learning;
Including,
The learning difficulty of the learning data set acquired through active learning is,
Distinguished depending on the time of acquisition of data through active learning,
The second learning data set has a higher learning difficulty compared to the first training data set,
Including a data set recently acquired compared to the first learning data set in the process of active learning,
The learning difficulty level differentiated according to the acquisition period is,
The neural network model is used in a curriculum-based learning process that performs learning while gradually increasing the learning difficulty,
The neural network model is,
Learned based on a first loss function applied to learning based on the first learning data set and a second loss function applied to learning using the second learning data set,
The first loss function is,
a first sub-loss function for learning the neural network model based on the first training data set; and
a second sub-loss function for distilling knowledge about the neural network model, which is performed based on a neural network model previously learned through active learning;
Including,
Computer-readable recording medium.

Images