KR102392383B1 - Method and apparatus for transmitting video data - Google Patents

Abstract

A video data transmission method performed by a computing device including at least one processor according to an embodiment of the present disclosure is disclosed. The method may include the steps of: obtaining a user preference including a weight for each of two or more quality indicators indicating video quality; and determining a data transmission amount per unit time to allow the integrated quality reflecting the acquired user preference to reach a target quality through a reinforcement learning decision model.

Description

Video data transmission method and device {METHOD AND APPARATUS FOR TRANSMITTING VIDEO DATA}

The present invention relates to a method for transmitting video data, and more particularly, to a method for transmitting video data based on reinforcement learning using a neural network.

In the existing industry, as one of the methods for providing a video, an 'Adaptive Bitrate Streaming' method has been used. Adaptive bitrate streaming is a technology that transmits a piece of video data with the highest quality at a level that bandwidth can digest based on the user's network environment and computing environment. That is, it refers to a technology for encoding a video to be transmitted at various bit rates, and then adjusting the bit rate for each piece of video data according to the user's situation and transmitting it. 1 is an exemplary diagram illustrating an adaptive bitrate streaming method for adjusting a bitrate of content.

However, the existing video data transmission method such as adaptive bitrate streaming has a problem in that it cannot provide an optimal bitrate in consideration of user preferences. Most of the existing methods only transmit video data according to a user's designation of a bit rate value, so there has been a demand in the art for an optimal video data transmission method in consideration of the user's preference. In addition, there has been a demand for an optimal video data transmission method suitable for not only a preference for one element, such as a bitrate value, but also a multidimensional preference for various elements.

Korean Patent Registration "KR1840685" discloses 'a method and apparatus for returning a guaranteed bit rate for variable bit rate media transmission'.

The present disclosure has been devised in response to the above background art, and an object of the present disclosure is to provide an efficient video data transmission method.

Disclosed is a video data transmission method performed by a computing device including at least one processor according to an embodiment of the present disclosure for realizing the above-described problem. The method includes: obtaining a user preference including a weight for each of two or more quality indicators indicating video quality; and determining a data transmission amount per unit time for the integrated quality reflecting the acquired user preference to reach a target quality through a reinforcement learning decision model.

In an alternative embodiment, the two or more quality indicators include a bitrate value, a framerate value, a rebuffering time value indicating the time required to fill the buffer, and video data transmitted from the server. At least one of a delay time value indicating a difference between the received time and the time at which the client receives the moving picture data may be included.

In an alternative embodiment, the integrated quality may be calculated based on a result of an operation between a weight vector indicating the user preference and a quality vector indicating video quality.

In an alternative embodiment, the reinforcement learning decision model comprises an artificial neural network layer comprising at least one node, and the learning method of the reinforcement learning decision model comprises: wherein the reinforcement learning decision model relates to video quality from an Environment. obtaining status information; obtaining, by the reinforcement learning decision model, a user preference including a weight for each of two or more quality indicators; determining, by the reinforcement learning decision model, a data transmission amount per unit time based on the state information and the user preference; and obtaining, by the reinforcement learning decision model, a reward from the environment.

In an alternative embodiment, the reinforcement learning decision model may determine, based on a Pareto optimal solution, a data transmission amount per unit time for the integrated quality reflecting the obtained user preference to reach a target quality.

In an alternative embodiment, when the preference weight value for at least one of the preference weight values for each of the two or more quality indicators included in the user preference is not obtained, the user preference including at least one pre-entered user preference The method may further include obtaining user preference distribution data based on the set.

In an alternative embodiment, the user preference distribution data may be obtained based on the user preference set through a semi-supervised learning prediction model.

In an alternative embodiment, after dividing the user preference set into N predetermined preference subsets based on the obtained user preference distribution data, a first user preference for a first preference subset and a second preference subset for the first preference subset calculating a second user preference for calculating a data transmission amount per first candidate unit time so that the integrated quality reflecting the first user preference reaches a target quality through a reinforcement learning decision model; calculating a data transmission amount per second candidate unit time so that the integrated quality reflecting the second user preference reaches a target quality through a reinforcement learning decision model; and determining a final data transmission amount per unit time based on the first candidate data transmission amount per unit time and the second candidate data transmission amount per unit time.

A computer program stored in a computer-readable storage medium is disclosed according to an embodiment of the present disclosure for realizing the above-described problems. When the computer program is executed on one or more processors, it causes the following operations to be performed for transmitting video data, wherein the operations include: a user preference (user preference) including a weight for each of two or more quality indicators indicating video quality preference); and determining a data transmission amount per unit time so that the integrated quality in which the acquired user preference is reflected through the reinforcement learning decision model reaches a target quality.

Disclosed is an apparatus for transmitting video data according to an embodiment of the present disclosure for realizing the above-described problems. The apparatus may include one or more processors; Memory; and a network unit, wherein the at least one processor obtains user preference including a weight for each of two or more quality indicators indicating video quality, and obtains a user preference through a reinforcement learning decision model. It is possible to determine the amount of data transmission per unit time for the integrated quality reflecting the preference to reach the target quality.

The present disclosure may provide an efficient method for transmitting video data through a reinforcement learned network function.

1 is an exemplary diagram illustrating an adaptive bitrate streaming method for adjusting a bitrate of content.
2 is a block diagram of a computing device for transmitting video data according to an embodiment of the present disclosure.
3 is a schematic diagram illustrating a network function according to an embodiment of the present disclosure;
4 is a conceptual diagram illustrating an agent and an environment for explaining a reinforcement learning method of a reinforcement learning decision model according to an embodiment of the present disclosure.
5 is a conceptual diagram exemplarily illustrating an agent and an environment in order to explain a reinforcement learning method of a reinforcement learning decision model according to another embodiment of the present disclosure.
6 is a graph for explaining a Pareto optimal solution according to an embodiment of the present disclosure.
7 is a flowchart illustrating a process in which a computing device transmits video data using a reinforcement learning decision model according to an embodiment of the present disclosure.
8 is a flowchart illustrating a process in which a computing device transmits video data using a reinforcement learning decision model according to another embodiment of the present disclosure.
9 is a simplified, 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 descriptions.

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 execution of software. For example, a component can be, but is not limited to being, 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 may 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. In addition, these components can execute from various computer readable media having various data structures stored therein. Components may communicate via a network such as the Internet with another system, for example, via a signal having one or more data packets (eg, data and/or signals from one component interacting with another component in a local system, distributed system, etc.) may communicate via local and/or remote processes depending on the data being transmitted).

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 context, "X employs A or B" is intended to mean one of the natural implicit substitutions. That is, X employs A; X employs B; or when X employs both A and B, "X employs A or B" may apply to either of these cases. It should also be understood that the term “and/or” as used herein refers to and includes 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 feature and/or element in question is 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 unless it is clear from context to refer to a singular form, the singular in the specification and claims should generally be construed to mean “one or more”.

And, the term “at least one of A or B” should be interpreted as meaning “includes only A”, “includes only B”, and “combined with the configuration of A and B”.

Those skilled in the art will further appreciate that the various illustrative logical blocks, configurations, modules, circuits, means, logics, and algorithm steps described in connection with the embodiments disclosed herein may be implemented in electronic hardware, computer software, or combinations of both. It should be recognized that they can be implemented with To clearly illustrate this 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 as hardware or software depends upon 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 the present disclosure.

Descriptions of the presented embodiments are provided to enable those 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 of the present disclosure. The generic principles defined herein may be applied to other embodiments without departing from the scope of the present disclosure. Therefore, the present invention is not limited to the embodiments presented herein. This invention is to be interpreted in its widest scope consistent with the principles and novel features presented herein.

2 is a block diagram of a computing device for transmitting video data according to an embodiment of the present disclosure.

The configuration of the computing device 100 shown in FIG. 2 is only a simplified example. In an embodiment of the present disclosure, the computing device 100 may include other 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 include one or more cores, and a central processing unit (CPU) of a computing device, a general purpose graphics processing unit (GPGPU), and a tensor processing unit (TPU). unit) and the like, and may include a processor for data analysis and deep learning. The processor 110 may read a computer program stored in the memory 130 to 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 the neural network. The processor 110 for learning of the neural network, such as processing input data for learning in deep learning (DL), extracting features from the input data, calculating an error, updating the weight of the neural network using backpropagation calculations can be performed. At least one of a CPU, a GPGPU, and a TPU of the processor 110 may process learning of a network function. For example, the CPU and the GPGPU can process learning of a network function and data classification using the network function. Also, in an embodiment of the present disclosure, learning of a network function and data classification using the network function may be processed by using the processors of a plurality of computing devices together. In addition, the computer program executed in the computing device according to an embodiment of the present disclosure may be a CPU, GPGPU or TPU executable program.

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, a hard disk type, a multimedia card micro type, or a card type memory (eg, SD or XD memory, etc.), Random Access Memory (RAM), Static Random Access Memory (SRAM), Read-Only Memory (ROM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Programmable Read (PROM) -Only Memory), a magnetic memory, a magnetic disk, and an optical disk may include at least one type of storage medium. The computing device 100 may operate in relation to a web storage that performs a storage function of the memory 130 on the Internet. The description of the above-described memory is only an example, and the present disclosure is not limited thereto.

The network unit 150 according to an embodiment of the present disclosure may include any wired/wireless communication network capable of transmitting and receiving any type of data and signals.

In the present disclosure, the network unit 150 may be configured regardless of its communication mode, such as wired and wireless, and may be configured with various communication networks such as a personal area network (PAN) and a wide area network (WAN). can In addition, the network may be a well-known World Wide Web (WWW), and may use a wireless transmission technology used for short-range communication, such as infrared (IrDA) or Bluetooth (Bluetooth).

The techniques described herein may be used in the networks mentioned above, as well as in other networks.

In the present disclosure, the video data transmission method performed by the computing device 100 includes the steps of obtaining user preference including weights for each of two or more quality indicators indicating video quality and a reinforcement learning decision model and determining a data transmission amount per unit time so that the acquired integrated quality reflecting the user preference reaches a target quality through .

In the present disclosure, 'data transmission amount per unit time' means the size per unit time of video data transmitted by a subject such as a server or uploader capable of providing a video service. The data transmission amount per unit time may be expressed as a value having units such as, for example, the number of bits per second (bps, bit per second) and the number of frames per second. In the present disclosure, 'data transmission amount per unit time' may be used interchangeably with 'bitrate'.

In the present disclosure, 'video quality' may be expressed by a plurality of quality indicators. Two or more quality indicators indicating video quality include a bitrate value, a framerate value, a rebuffering time value indicating the time required to fill the buffer, and the time the video data is transmitted from the server. At least one of a delay time value indicating a difference in time at which the client receives the moving picture data may be included. A bitrate value refers to the number of bits that a computing device for transmitting or receiving video data processes for each specific unit time. The frame rate value means the number of frames processed per unit time in video data. The rebuffering time value means a time for the computing device 100 to fill the buffer with video data in order to reproduce the video. The delay time value represents a difference between the time the video data is transmitted from the server and the time the client receives the video data, and may be a value reflecting the state of the network. In embodiments of the present disclosure, two or more quality indicators indicating video quality may include other components for quantitatively measuring video quality, and only some of the disclosed components constitute quality indicators for indicating video quality. It will be apparent to those skilled in the art that this may be the case. As described above, as the value of at least some of the two or more quality indicators indicating video quality increases, the 'QoE (Quality of Experience)' for the video service may increase. For example, in the case of a bit rate value or a frame rate value, as the value increases, the video quality improves, so QoE may increase. Conversely, at least some of the quality indicators may decrease QoE as their values increase. Typically, when the absolute value of the rebuffering time value or the delay time value increases, the waiting time increases, so the user's QoE is reduced. In the present disclosure, the term 'video quality' may be used interchangeably with 'quality of experience', 'QoE', or 'quality of service'.

The user preference of the present disclosure may include a preferred weight for each of two or more quality indicators indicating video quality. The preference weight for each quality indicator included in the user preference may be interpreted as meaning the importance of the corresponding quality indicator according to the user. The ratio of the preference weight for each quality indicator may be interpreted as meaning a ratio of the importance of each quality indicator according to the user. The user preference may be expressed in the form of a vector of preference weights for each of the two or more quality indicators. The vector representation of the user preference may be interchangeably referred to as a 'preference weight vector'. In one embodiment, when user A inputs preference weights of bit rate and delay time values as '10' and '1', respectively, user preference for [bit rate, delay time value] is '(10,1) It can be expressed as a preferred weight vector such as '. In addition, in the above embodiment, the preferred weight vector (that is, (10,1)) is “User A receives a high-quality video by increasing the 'bit rate' rather than receiving a faster video by reducing the 'delay time value' by 10 It has the importance or preference of the ship.” User preference according to the present disclosure means weights or values assigned by a user to each of the quality indicators for the purpose of optimizing the quality of experience.

The integrated quality according to the present disclosure may be calculated based on an operation result between a preference weight vector indicating user preference and a quality vector indicating video quality.

In the present disclosure, “Unified QoE” may be used as a term meaning “quality information in which both video quality (ie, QoE) and user preference are considered”. In general, when transmitting video data, there is no single solution that maximizes all quality indicators of video quality of experience (QoE) due to the limitation of computing resources. For example, in a state where network resources are limited, if a user considers only the video quality and sets the preferred weight for the bit rate to the maximum, a relatively long refurbishing time or a long delay time due to a high bit rate value have to endure Conversely, if the user requests fast video transmission and sets the preferred weight for the delay time to the maximum (that is, to give a high weight to reducing the delay time), the video quality will be deteriorated due to the low bitrate value. . As in the above example, there is no unique solution that simultaneously satisfies all conditions for a plurality of quality indicators (eg, bit rate, delay time, etc.) included in quality of experience (QoE). In other words, there may not be a unique solution that satisfies all conditions at the same time in the competitive conditions in which one has to give up the other index due to the limitation of computing resources in order to improve one index. Accordingly, the present invention discloses 'integrated quality' in which the user's preference is reflected. In the present disclosure, the term “target quality” may be used interchangeably with “integrated quality value in the case where the value of the calculated integrated quality is the maximum for a given user preference”. The formula for the integrated quality can be expressed as Equation (1).

[Equation 1]

는 동영상 품질 벡터를 구성하는 구성요소 중 이익 벡터(benefit vector)를 나타낸다. 상기 이익 벡터

는 동영상 품질을 구성하는 품질 지표들 중 값이 증가할수록 동영상 품질에 긍정적인 영향을 미치는 지표들(예를 들어, 비트레이트, 프레임레이트 등)을 벡터의 성분으로서 포함할 수 있다. 반면

는 동영상 품질 벡터를 구성하는 구성요소 중 페널티 벡터(penalty vector)를 나타낸다. 상기 페널티 벡터

는 동영상 품질을 구성하는 품질 지표들 중 값이 증가할수록 동영상 품질에 부정적인 영향을 미치는 지표들(예를 들어, 리퍼버링 시간, 딜레이 시간 등)을 벡터의 성분으로서 포함할 수 있다.

는 선호 가중치 벡터에 포함된 적어도 하나의 선호 가중치 중에서 이익 벡터에 대응되는 선호 가중치들을 포함하는 부분 가중치 벡터를 나타낸다.

는 선호 가중치 벡터에 포함된 적어도 하나의 선호 가중치 중에서 페널티 벡터에 대응되는 선호 가중치들을 포함하는 부분 가중치 벡터를 나타낸다.

는 이익 벡터에 대한 비선형적 선호도를 표현하기 위한 지수 가중치를 나타낸다.

는 예를 들어 1, 2.5 등의 실수값을 가질 수 있다.

는 선호 가중치 벡터의 구성 성분 중에서 이익 벡터에 대응되는 가중치 벡터의 원소들 총합을 나타낸다.

는 선호 가중치 벡터의 구성 성분 중에서 페널티 벡터에 대응되는 가중치 벡터의 원소들 총합을 나타낸다. 본 개시에서 사용자 선호도를 표현하기 위한 선호 가중치 벡터는 [

]와 같이 표현될 수 있다. 사용자는 상기 선호 가중치 벡터에 포함된 어떠한 인자라도 언제든지 조절할 수 있다. 본 개시에 있어서 타겟 품질은 상기한 선호 가중치 벡터에 대해서 상기 수학식 1로 표현되는 통합 품질이 최대인 경우의 값을 의미하며, 본 개시에 따른 컴퓨팅 장치(100)는 이러한 타겟 품질을 만족시키기 위해 강화학습 결정 모델을 이용한다. 강화학습 결정 모델에 관하여는 후술하여 자세히 설명한다. 전술한 통합 품질과 관련한 수학식 1은 일 예시에 불과할 뿐, 본 개시는 사용자 선호도와 동영상 품질을 각각 벡터로 표현하고 벡터 사이의 연산 결과에 기초하여 통합 품질을 산출할 수 있는 방법을 제한없이 포함할 수 있다.The present disclosure understands a moving picture quality (QoE) vector as two components through Equation (1).

represents a benefit vector among the components constituting the video quality vector. said profit vector

may include indicators (eg, bit rate, frame rate, etc.) that positively affect the video quality as the value of the quality indicators constituting the video quality increases as a component of the vector. On the other hand

represents a penalty vector among components constituting the video quality vector. said penalty vector

may include, as components of a vector, indicators that negatively affect the video quality (eg, repurposing time, delay time, etc.) as the value of the quality indicators constituting the video quality increases.

denotes a partial weight vector including preference weights corresponding to the profit vector among at least one preference weight included in the preference weight vector.

denotes a partial weight vector including preference weights corresponding to the penalty vector among at least one preference weight included in the preference weight vector.

denotes an exponential weight for expressing a non-linear preference for a profit vector.

may have a real value such as 1, 2.5, etc., for example.

represents the sum of the elements of the weight vector corresponding to the profit vector among the constituents of the preference weight vector.

denotes the sum of elements of the weight vector corresponding to the penalty vector among constituents of the preference weight vector. In the present disclosure, the preference weight vector for expressing user preference is [

] can be expressed as The user can adjust any factor included in the preference weight vector at any time. In the present disclosure, the target quality means a value when the integration quality expressed by Equation 1 is the maximum with respect to the preference weight vector, and the computing device 100 according to the present disclosure is used to satisfy this target quality. Reinforcement learning decision model is used. The reinforcement learning decision model will be described later in detail. Equation 1 related to the above-described integrated quality is only an example, and the present disclosure includes, without limitation, a method for expressing user preference and video quality as vectors and calculating integrated quality based on the result of calculation between the vectors. can do.

When the integrated quality is expressed as in Equation 1 according to the present disclosure, the computing device 100 selects one factor that positively affects video quality as the value increases and one factor that negatively affects video quality as the value increases. It is possible to obtain a scalarized integrated quality value by calculating along with user preference through the formula of . Accordingly, in the video data transmission method according to the present disclosure, the computing device 100 defines an integrated quality value according to user preference and determines the data transmission amount per unit time based thereon.

The computing device 100 according to the present disclosure may determine the data transmission amount per unit time for the integrated quality reflecting user preference to reach the target quality as described above through the reinforcement learning decision model.

3 is a schematic diagram illustrating a network function according to an embodiment of the present disclosure; The reinforcement learning decision model according to the present disclosure may include at least one or more nodes. The reinforcement learning decision model according to the present disclosure may include a neural network structure including one or more neural network layers.

Throughout this specification, the terms model, decision model, predictive model, neural network, network function, and neural network may be used interchangeably. A neural network may be composed of a set of interconnected computational units, which may generally be referred to as nodes. These nodes may also be referred to as neurons. The neural network is configured to include at least one node. Nodes (or neurons) constituting the neural networks may be interconnected by one or more links.

In the neural network, one or more nodes connected through a link may relatively form a relationship between an input node and an output node. The concepts of an input node and an output node are relative, and any node in an output node relationship with respect to one node may be in an input node relationship in a relationship with another node, and vice versa. As described above, an input node-to-output node relationship may be created around a link. One or more output nodes may be connected to one input node through a link, and vice versa.

In the relationship between the input node and the output node connected through one link, the value of the data of the output node may be determined based on data input to the input node. Here, a link interconnecting the input node and the output node may have a weight. The weight may be variable, and may be changed by a user or an algorithm in order for the neural network to perform a desired function. For example, when one or more input nodes are interconnected to one output node by respective links, the output node sets values input to input nodes connected to the output node and links corresponding to the respective input nodes. An output node value may 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 an output node relationship in the neural network. The characteristics of the neural network may be determined according to the number of nodes and links in the neural network, the correlation between the nodes and the links, and the value of a weight assigned to each of the links. For example, when the same number of nodes and links exist and there are two neural networks having different weight values of the links, 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 constituting the neural network may constitute a layer. Some of the nodes constituting the neural network may configure one layer based on distances from the initial input node. For example, a set of nodes having a distance of n from the initial input node may constitute n layers. The distance from the initial input node may be defined by the minimum number of links that must be passed to reach the corresponding node from the initial input node. However, the definition of such a layer is arbitrary for description, and the order of the layer in the neural network may be defined in a different way from the above. For example, a layer of nodes may be defined by a distance from the final output node.

The initial input node may mean one or more nodes to which data is directly input without going through a link in a relationship with other nodes among nodes in the neural network. Alternatively, in a relationship between nodes based on a link in a neural network, it may mean nodes that do not have other input nodes connected by a link. 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. In addition, the hidden node may mean nodes constituting the neural network other than the first input node and the last output node.

The neural network according to an embodiment of the present disclosure may be 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 progresses from the input layer to the hidden layer. can Also, in the neural network according to another embodiment of the present disclosure, 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 may be reduced as the number of nodes 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 the number of nodes progresses from the input layer to the hidden layer. can The neural network according to another embodiment of the present disclosure may be a neural network in a combined form 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 be used to identify the latent structures of data. In other words, it can identify the potential structure of photos, texts, videos, voices, and music (e.g., what objects are in the photos, what the text and emotions are, what the texts and emotions are, etc.) . Deep neural networks include convolutional neural networks (CNNs), recurrent neural networks (RNNs), auto encoders, generative adversarial networks (GANs), and restricted boltzmann machines (RBMs). machine), a deep belief network (DBN), a Q network, a U network, a Siamese network, and a Generative Adversarial Network (GAN). The description of the deep neural network described above is only an example, and the present disclosure is not limited thereto.

In an embodiment of the present disclosure, the network function may include an autoencoder. The auto-encoder may be a kind of artificial neural network for outputting output data similar to input data. The auto encoder may include at least one hidden layer, and an odd number of hidden layers may be disposed between the input/output layers. The number of nodes in 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 reduction from the bottleneck layer to the output layer (symmetrically with the input layer). The auto-encoder can perform non-linear dimensionality reduction. The number of input layers and output layers may correspond to a dimension after preprocessing the input data. In the auto-encoder structure, the number of nodes of the hidden layer included in the encoder may have a structure that decreases 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 a certain number or more (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, semi-supervised learning, and reinforcement learning. Learning of the neural network may be a process of applying knowledge for the neural network to perform a specific operation to the neural network.

A neural network can be trained in a way that minimizes output errors. In the training of a neural network, iteratively input the training data into the neural network, calculate the output of the neural network and the target error for the training data, and calculate the error of the neural network from the output layer of the neural network to the input layer in the direction to reduce the error. It is a process of updating the weight of each node in the neural network by backpropagation in the direction. In the case of teacher learning, learning data in which the correct answer is labeled in each learning data is used (ie, labeled learning data), and in the case of comparative learning, the correct answer may not be labeled in each learning data. That is, for example, learning data in the case of teacher learning related to data classification may be data in which categories are labeled in each of the learning data. Labeled training data is input to the neural network, and an 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 comparison learning related to data classification, an error may be calculated by comparing the input training data with the neural network output. The calculated error is back propagated in the 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 change amount of the connection weight of each node to be updated may be determined according to a learning rate. The computation of the neural network on the input data and the backpropagation of errors can constitute a learning cycle (epoch). The learning rate may be applied differently according to the number of repetitions of the learning cycle of the neural network. For example, in the early stage of learning of a neural network, a high learning rate can be used to enable the neural network to quickly obtain a certain level of performance, thereby increasing efficiency, and using a low learning rate at a later stage of learning can increase accuracy.

In the training of neural networks, in general, the training data may be a subset of real data (that is, data to be processed using the trained neural network), and thus, the error on the training data is reduced, but the error on the real data is reduced. There may be increasing learning cycles. Overfitting is a phenomenon in which errors on actual data increase by over-learning on training data as described above. For example, a phenomenon in which a neural network that has learned a cat by seeing 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 errors in machine learning algorithms. In order to prevent such overfitting, various optimization methods can be used. In order to prevent overfitting, methods such as increasing the training data, regularization, and dropout that deactivate some of the nodes of the network in the process of learning, and the use of a batch normalization layer are applied. can

The computing device 100 according to the present disclosure may train a reinforcement learning decision model for determining the data transmission amount per unit time based on the reinforcement learning method. The reinforcement learning decision model including at least one artificial neural network layer may update a parameter or bias value regarding at least one node included in the artificial neural network layer in the reinforcement learning process. Hereinafter, a learning method of the reinforcement learning decision model according to the present disclosure will be described with reference to FIGS. 4 to 5 .

)에 기초하여 가능한 행동들의 확률 분포에 따라 임의의 행동(

)을 결정하고, 환경(430)으로부터 갱신된 상태 정보(

)와 보상(

)을 받는다. 이러한 상호 작용에 기반하여 에이전트(410)는 주어진 환경(430)에서 리턴을 최대화하는 정책(Policy)을 학습한다. 상기 정책이란 에이전트(410)가 특정 상태에 대해 특정 행동을 할 확률에 관한 집합을 의미할 수 있다. 본 개시에 있어서 환경(430)은 컴퓨팅 장치(100) 내부 또는 외부에 존재하여 강화학습 결정 모델의 연산과정과 별도로 상태 정보 및 보상을 산출할 수도 있다. 4 is a conceptual diagram illustrating an agent and an environment to explain a reinforcement learning method of a reinforcement learning decision model according to an embodiment of the present disclosure. Reinforcement learning is a method in which an artificial neural network model selects a behavior and trains an artificial neural network model based on a reward given for the selected behavior. The reward given to the artificial neural network model in the reinforcement learning process may be a reward accumulated by the results of several actions. Reinforcement learning creates an artificial neural network model that maximizes reward or return by considering various states and rewards according to actions through learning. The return (return) may be used interchangeably with "cumulative reward" and has the same meaning. In the present disclosure, the reinforcement learning decision model can be used interchangeably with an agent as a subject that determines a behavior. In the present disclosure, an environment (Environment, Env) may be used as a concept corresponding to the agent 410 . The environment 430 may provide the agent 410 with state information that may be a basis for the agent 410 to determine an action. The agent 410 may then determine an action based on the state information obtained from the environment 430 . When the agent 410 transfers the determined action to the environment 430 , the agent 410 may receive a reward based on the action and next state information from the environment 430 . Reinforcement learning when the environment 430 can know the reward function as a criterion for determining the reward and the transition probability distribution function as the criterion for determining the next state information after the environment 430 receives the action from the agent 410 is called "model-based reinforcement learning. On the other hand, when the agent 410 cannot know the reward function of the environment 430 and the transition probability distribution function of the environment 430, reinforcement learning is "model-free" (Model-free) is called "reinforcement learning. When state information and next state information or state information and updated state information are expressed in relation to time t, at any time t, the agent 410 is Acquired status information (

) based on a random action (

), and updated state information from the environment 430 (

) and reward (

) is received. Based on this interaction, the agent 410 learns a policy that maximizes the return in a given environment 430 . The policy may mean a set regarding the probability that the agent 410 will perform a specific action for a specific state. In the present disclosure, the environment 430 may exist inside or outside the computing device 100 to calculate state information and rewards separately from the operation process of the reinforcement learning decision model.

In the present disclosure, the method of the agent determining the behavior is, for example, at least one of a value-based behavior determination method, a policy-based behavior determination method, and a behavior determination method based on both values and policies. can be based on The value-based action determination method is a method of determining an action giving the highest value in each state based on a value function. Examples of the value-based behavior determination method may include Q-learning, Deep Q-Network (DQN), and the like. The policy-based action determining method is a method of determining an action based on a final return and a policy function without a value function. An example of the policy-based action determination method may include a policy gradient technique. The behavior determination method based on both the value and the policy is a method of determining the agent's behavior by learning in a manner that the value function evaluates the behavior when the policy function determines the behavior. Methods for determining actions based on both values and policies may include, for example, the Soft Actor-Critic algorithm. For a description of the specific content related to the learning method of the reinforcement learning model, the nature paper "Human-level control through deep reinforcement learning" (published date: February 25, 2015, author: Mnih, V., Kavukcuoglu, K., Silver, D. et al.).

5 is a conceptual diagram exemplarily illustrating an agent and an environment in order to explain a reinforcement learning method of a reinforcement learning decision model according to another embodiment of the present disclosure. Hereinafter, content overlapping with the embodiment described above with respect to FIG. 4 will be omitted, and differences will be mainly described. Reference numeral 510 denotes an agent based on a reinforcement learning decision model. The computing device 100 determines the data transmission amount per unit time through the agent 510 . Reference numeral 530 denotes an environment for notifying state information to the reinforcement learning decision model. The environment 530 of FIG. 5 may transmit the user preference W as well as the state information to the agent 510 together. Reference numeral 550 denotes a user inputting user preference, and the user preference may be input from the user 550 as an arbitrary value at any time. In the present disclosure, the state information transmitted from the environment 530 to the agent 510 may include information about video quality. In general, the state information means a value that the agent acquires from the environment according to the agent's behavior in each step. However, in the present disclosure, user preference is a value randomly given by the user 550, not a value observed or obtained by the behavior of the model, and may be constant regardless of the agent's behavior for each step, and irrespective of the progress of the step. It is a value that can be changed in a specific step. That is, in the present disclosure, user preferences have properties that are essentially different from status information. Specifically, for example, in transmitting video data, when the agent determines the data transmission amount per unit time for each step, the network status, transmission speed, bit rate, delay time, etc. can be included in the status information that can be checked at every step. there is. On the other hand, the user preference indicating the preference between the 'video quality' and the 'delay time' of the user 550 may be changed regardless of the progress of each step by the agent. As described above, the present disclosure discloses state information and user preference as separate concepts for learning a decision model that determines a bit rate based on user preference. transmission method can be achieved.

The status information obtained by the agent 510 from the environment 530 according to the present disclosure may include quality indicators related to video quality. The status information includes “timestamp of current chunk download”, “startup delay”, “throughput for past N chunks”, “length of remaining chunks” (remaining chunk length)”, “returned buffer size”, “the last chunk bitrate”, “sending bitrate of the last frame”, “round trip delay time (RTT)”, “latency gradient”, “packet loss rate” and “receiving bitrate of the last frame” At least one of them may be included. The user preference obtained by the agent 510 from the environment 530 according to the present disclosure may include a preference weight for at least one or more quality indicators among the above-described state information. The agent 510 according to the present disclosure may determine the data transmission amount per unit time in consideration of status information and user preference based on Equation 1 described above.

The reinforcement learning decision model according to the present disclosure may include an artificial neural network layer including at least one node. A method of learning a reinforcement learning decision model according to the present disclosure, the reinforcement learning decision model acquiring state information related to video quality from an environment; obtaining, by the reinforcement learning decision model, a user preference including a preference weight for each of two or more quality indicators; determining, by a reinforcement learning decision model, a data transmission amount per unit time based on the state information and the user preference; and the reinforcement learning decision model obtaining a reward from the environment.

The reinforcement learning decision model according to the present disclosure may be learned based on a plurality of episodes. The episode may refer to a sequence of (state, action, reward) from an initial state to a terminal state. The final state may be derived when a preset termination condition is satisfied or may be derived when a step of a preset size is performed. The step refers to at least one operation unit in which the reinforcement learning decision model receives a state, determines a behavior, and receives a reward or updated state information for the behavior. One episode may consist of a predetermined number of steps.

), 사용자 선호도를 나타내는 선호 가중치 벡터(

), 강화학습 결정 모델이 상기 상태 정보에 기초하여 결정하는 행동(

), 강화학습 결정 모델이 상기 행동의 결과로서 환경으로부터 획득한 보상(

)을 (상태 정보(

), 선호 가중치 벡터(

), 행동(

), 보상(

))의 순서쌍 형태로 학습 데이터로서 메모리(130)에 저장할 수 있다. 본 개시에 있어서 상기 강화학습 결정 모델이 상기 상태 정보 및 사용자 선호도에 기초하여 결정하는 행동은 다음 스텝에서의 단위 시간당 데이터 전송량을 포함할 수 있다. 상기 학습 데이터의 시점은, 시점 t에서 강화학습 결정 모델이 행동을 결정하고 결정한 행동의 결과로서 환경으로부터 갱신된 상태 정보 및 사용자 선호도를 획득하는 경우에 다음 상태에 대한 시점 t+1로 진행될 수 있다. 이 때 사용자 선호도는 환경으로부터 수신되기는 하나, 강화학습 결정 모델의 행동 결정과 무관하게 임의의 시점에 변경될 수 있다. 전술한 학습 데이터의 형태, 시점에 관한 서술은 일 예시에 불과하며 본 개시를 제한하지 않는다.In the reinforcement learning process of the model according to the present disclosure, the computing device 100 may acquire learning data for each step included in at least one episode. For example, the computing device 100 provides state information (

), a preference weight vector representing user preference (

), the behavior that the reinforcement learning decision model determines based on the state information (

), the reward that the reinforcement learning decision model obtains from the environment as a result of the action (

) to (status information (

), the preferred weight vector (

), action(

), compensation(

)) may be stored in the memory 130 as learning data in the form of an ordered pair. In the present disclosure, the action determined by the reinforcement learning decision model based on the state information and user preference may include a data transmission amount per unit time in the next step. The time point of the learning data may proceed to time point t+1 for the next state when the reinforcement learning decision model determines a behavior at time t and obtains updated state information and user preference from the environment as a result of the determined behavior. . At this time, although the user preference is received from the environment, it may be changed at any time regardless of the action decision of the reinforcement learning decision model. The description of the form and timing of the above-described learning data is merely an example and does not limit the present disclosure.

The process of the computing device 100 learning the reinforcement learning decision model according to the present disclosure may include a process of correcting a weight or bias value of each node included in the reinforcement learning decision model. The process of correcting the weight or bias value of each node included in the reinforcement learning decision model may be based on the backpropagation technique as described above. In a specific embodiment, when the reward included in the training data for learning the reinforcement learning decision model is positive, the absolute value of the weight or bias value of at least one node included in the reinforcement learning decision model is increased to calculate the corresponding reward. can be Conversely, when the reward included in the training data for learning the reinforcement learning decision model is negative, the absolute value of the weight or bias value of at least one node involved in calculating the corresponding reward may be reduced. The above-described learning process of the reinforcement learning decision model is merely a description for illustration and does not limit the present disclosure.

The reinforcement learning decision model according to the present disclosure may determine, based on the Pareto optimal solution, the amount of data transmission per unit time for the integrated quality reflecting the acquired user preference to reach the target quality. The 'Pareto optimum' refers to a state in which it is impossible to create a more beneficial state for one factor without damaging another factor in a resource allocation state. The computing device 100 according to the present disclosure may learn an optimal policy for reaching at least one point of a Pareto optimal curve for the reinforcement learning decision model. The optimal policy may mean information about the behavior that the reinforcement learning decision model should select in each state in order to maximize the sum of returns. In this regard, it will be described in detail with reference to FIG. 6 .

6 is a graph for explaining a Pareto optimal solution according to an embodiment of the present disclosure. In an embodiment of the present disclosure, the X-axis may indicate video quality. In this case, the Y-axis may indicate the degree of delay time reduction. In this embodiment, it can be understood that the image quality increases as the X-axis coordinate increases, and the delay time decreases as the Y-axis coordinate increases. Point A is any point located closer to the origin than the Pareto optimal curve 650 . Point A is not “Pareto-optimal” in that there is room for maintaining or increasing the video quality while further reducing the delay time, and also maintaining or reducing the delay time while further increasing the video quality. Points B and C represent any two points on the Pareto optimal curve 650 . Therefore, both points B and C represent optimal solutions for specific user preferences.

Hereinafter, in the present disclosure, a method for the reinforcement learning decision model to search for an optimal solution corresponding to different user preferences based on a Pareto optimal solution will be described. As described above, the reinforcement learning decision model can learn a behavior policy that reaches an arbitrary point on the Pareto optimal curve 650 through the reinforcement learning process. At this time, if there is a behavior policy that can reach an arbitrary point on the Pareto optimal curve 650 , there must be an arbitrary preference weight vector that maximizes the reward according to the corresponding behavior policy and the result of the dot product operation. Under the above conditions, the reinforcement learning decision model according to the present disclosure learns a behavior policy that can reach an arbitrary point on the Pareto optimal curve 650, and then follows the behavior policy. It is possible to learn what the preferred weight vector value is. After learning is completed, the reinforcement learning decision model can find the optimal policy that maximizes the result of the dot product operation when the preference weight vector for user preference is given.

For example, in the Pareto optimal curve 650, if there is a behavior policy that can reach the point B state among the behavior policies by the reinforcement learning decision model, it is the optimal value for the preference weight vector 610 for the point B. It could be a policy of action. Also, if there is a behavior policy that can reach the point C state among the behavior policies based on the reinforcement learning decision model, this may be an optimal behavior policy for the preference weight vector 630 for the point C. More specifically, the ratio between the X coordinate 631 of the point C and the Y coordinate 633 of the point C may represent the ratio of the horizontal component and the vertical component of the vector 630 pointing to the point C. The computing device 100 according to the present disclosure sets the ratio of the horizontal component and the vertical component of the vector 630 pointing to the point C to the video quality and delay time reduction corresponding to the action policy for reaching the point C state among the action policies. It can be obtained as a ratio of user preference to

A detailed description of a method for searching for an optimal solution corresponding to different user preferences based on the Pareto optimal solution is provided in the paper "Computing convex coverage sets for faster multi-objective coordination" (published date of publication), which is incorporated herein by reference in its entirety. : 31 Mar 2015, Authors: Diederik Marijn Roijers, Shimon Whiteson, Frans A. Oliehoek).

The computing device 100 according to the present disclosure may obtain user preference by directly receiving a preference weight value for each of two or more quality indicators from a user. Conversely, the computing device 100 according to the present disclosure may fail to receive a clear input value from the user with respect to at least some of the plurality of quality indicators. At this time, when the preference weight value for at least one of the preference weight values for each of the two or more quality indicators included in the user preference is not obtained, the computing device 100 according to the present disclosure determines at least one previously input user preference. User preference distribution data may be obtained based on the user preference set including

The user preference set according to the present disclosure may refer to a set related to user preferences input by a past user. The user preference distribution data according to the present disclosure may refer to statistical data calculated based on at least one or more user preference data input in the past. According to an embodiment, the user preference distribution data may include at least one of an average value, a median value, and a quartile value of user preferences. In addition, the user preference distribution data may include at least one of a variance, a standard deviation, and a reference value for forming a predetermined N subsets of user preferences. The average value, the median value, the quartile value, etc. may be calculated for each component on the preference weight vector. The reference value for forming the predetermined N subsets may be a value existing for each vector component in order to divide each component of the vector in order to cluster the preference weight vector. A K-means clustering method may be used as a method of clustering the preferred weight vector.

The user preference distribution data according to the present disclosure may be obtained based on a user preference set through a semi-supervised learning prediction model. The prediction model may include at least one artificial neural network layer. According to the present disclosure, more accurate distribution data may be obtained by using a predictive model when obtaining user preference distribution data based on a user set. The predictive model may be trained by a semi-supervised learning method. Semi-supervised learning is a learning method that uses labeled data and unlabeled data together for training when labeled data is insufficient for training an artificial neural network model. The semi-supervised prediction model according to the present disclosure can generate more accurate distribution data even when the number of pre-input user preferences is insufficient. The semi-supervised learning method according to the present disclosure is the paper "Semi-supervised Deep Kernel Learning: Regression with Unlabeled Data by Minimizing Predictive Variance" (published date: May 26, 2018, author: Neal), which is incorporated herein by reference in its entirety Jean, Sang Michael Xie, Stefano Ermon).

In an embodiment of the present disclosure, when the computing device 100 obtains user preference distribution data, the computing device 100 may determine the data transmission amount per unit time through the following method using the reinforcement learning decision model. The method includes, after dividing the user preference set into N predetermined preference subsets based on the obtained user preference distribution data, a first user preference for a first preference subset and a second preference subset for a second preference subset It may include calculating user preference. In the predetermined N preference subset, predetermined N may be selected by a user as an arbitrary natural number. The preference subset means an individual preference set divided into N predetermined groups from among all previously input user preferences. The criterion for dividing the predetermined N groups may be such that each preference subset includes the same number of user preference data. The criterion for dividing the predetermined N groups may be a criterion for dividing a section occupied by the entire vector into N equal intervals regardless of the number of elements included in each preference subset. Specifically, the equal interval N equal division may be equal to N equal division of each component value included in the vector. In an embodiment for explanation, it is assumed that a plurality of user preferences included in the user preference set are divided into two preference subsets and divided at equal intervals for each vector component in the same way. If the first element value of the preference weight vector is distributed as a value between (0 to 10) and the second element value of the preference weight vector is distributed as a value between (0 to 1), in the above embodiment, the user preference set is the preference weight vector The first preference subset may be divided into a first preference subset consisting of preference weight vectors having a first element value of 5 or greater and a second element value of 0.5 or greater, and a second preference subset consisting of preference weight vectors other than the condition. It will be apparent to those skilled in the art that the above-described method for dividing the preference subset is merely an exemplary description. In the present disclosure, terms such as “first” or “second” that modify the preference subset above are only for distinguishing any two subsets of the preference subset in order, and do not limit the present disclosure. The computing device 100 may calculate the first user preference from the first preference subset. For example, a median value, an average value, etc. of preference weight vectors representing user preferences included in the first preference subset may be calculated as the first user preference. Similarly, the computing device 100 may calculate the second user preference for the second preference subset.

The computing device 100 according to the present disclosure may calculate the data transmission amount per first candidate unit time for the integrated quality reflecting the first user preference to reach the target quality through the reinforcement learning decision model. Also, the computing device 100 may calculate the data transmission amount per second candidate unit time for the integrated quality reflecting the second user preference to reach the target quality. In the present disclosure, the data transmission amount per first candidate unit time and the data transmission amount per second candidate unit time mean values calculated before the reinforcement learning decision model determines the final data transmission amount per unit time. As described above, in the present invention, even if user preferences are not explicitly given, data transmission amounts per a plurality of candidate units can be calculated from previously input user preferences based on user preference distribution data.

The computing device 100 according to the present disclosure may determine the final data transmission amount per unit time based on the first candidate data transmission amount per unit time and the second candidate data transmission amount per unit time. In the present disclosure, the final data transmission amount per unit time may be additionally determined based on the probability of the first user preference for the first preference subset and the probability of the second user preference with respect to the second preference subset. When the number of user preference elements included in the first preference subset is different from the number of user preference elements included in the second preference subset, the probability of the first user preference and the probability of the second user preference may have different values. The probability may mean an appearance frequency or a distribution probability value. The probability may be obtained from user preference distribution data for each preference subset.

Specifically, in determining the final data transmission amount per unit time, the computing device 100 selects the expected value for integrated quality when selecting the first candidate data transmission amount per unit time and the second candidate integrated quality when selecting the data transmission amount per unit time can be compared with each other. For example, the computing device 100 multiplies the distribution probability of the first user preference and the integrated quality value when selecting the data transmission amount per first candidate unit time for the first user preference, and the distribution probability of the second user preference and For the second user preference, the sum of the result of multiplying the integrated quality value when the data transmission amount per unit time of the first candidate is selected (the first final value), and the distribution probability of the first user preference and the second candidate for the first user preference ( The second final value) may be compared to determine the data transmission amount per final unit time. The above-described method of determining the data transmission amount per final unit time is merely an exemplary description, and the content of determining the final data transmission amount per unit time after calculating the expected value through a reinforcement learning decision model based on user preference distribution data is included without limitation. can do.

7 is a flowchart illustrating a process in which a computing device transmits video data using a reinforcement learning decision model according to an embodiment of the present disclosure. The computing device 100 may acquire ( 710 ) user preference including a preference weight for each of two or more quality indicators indicating video quality. The two or more quality indicators may mean quantitative values for representing video quality. The user preference may be expressed in the form of a preference weight vector when two or more preference weight values are included. The computing device 100 may determine ( 730 ) the amount of data transmission per unit time for the integrated quality reflecting the acquired user preference to reach the target quality through the reinforcement learning decision model. The reinforcement learning decision model may be a model including at least one artificial neural network node learned in a multi-objective markov decision process (MOMDP) situation for achieving a plurality of objectives. The reinforcement learning determination model may determine the data transmission amount per unit time for allowing the integrated quality value according to Equation 1, in which the user preference is reflected in the form of a preference weight vector, to reach the target quality value. The target quality may mean a value at which the value of the integrated quality is maximum in a given user preference.

8 is a flowchart illustrating a process in which a computing device transmits video data using a reinforcement learning decision model according to another embodiment of the present disclosure. The computing device 100 may acquire ( 810 ) user preference distribution data based on a user preference set including at least one previously input user preference. The user preference distribution data may be obtained through a semi-supervised prediction model. The user preference distribution data may include statistical data related to previously input user preferences. Computing device 100 divides the user preference set into predetermined N preference subsets based on the obtained user preference distribution data, and then sets the first user preference for the first preference subset and the second preference subset for the first preference subset. A second user preference may be calculated ( 830 ). The predetermined N preference subsets may be a result of dividing each component of a preference weight vector representing all user preferences included in the user preference set into N equally spaced sections. The computing device 100 calculates (850) the data transmission amount per first candidate unit time for the integrated quality reflecting the first user preference to reach the target quality through the reinforcement learning decision model, and through the reinforcement learning decision model A data transmission amount per second candidate unit time may be calculated ( 870 ) so that the integrated quality reflecting the second user preference reaches the target quality. Thereafter, the computing device 100 may determine ( 890 ) a final data transmission amount per unit time based on the first candidate data transmission amount per unit time and the second candidate data transmission amount per unit time. The final data transmission amount per unit time determination process may be additionally based on the expected value of the first user preference and the expected value of the second user preference.

9 is a simplified, general schematic diagram of an example computing environment in which embodiments of the present disclosure may be implemented. Although the present disclosure has been described above as being generally capable of being implemented by a computing device, those skilled in the art will appreciate that the present disclosure is a combination of hardware and software and/or in combination with computer-executable instructions and/or other program modules that may be executed on one or more computers. It will be appreciated that it can be implemented as a combination.

Generally, program modules include routines, programs, components, data structures, etc. that perform particular tasks or implement particular abstract data types. In addition, those skilled in the art will appreciate that the methods of the present disclosure can be applied 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. (each of which is It will be appreciated that other computer system configurations may be implemented, including those that may operate 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.

Computers typically include a variety of computer-readable media. Any medium accessible by a computer can be a computer-readable medium, and such computer-readable media includes volatile and nonvolatile media, transitory and non-transitory media, removable and non-transitory media. including 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 includes volatile and nonvolatile media, temporary and non-transitory media, 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. A computer-readable storage medium may be 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 the desired information.

Computer readable transmission media typically embodies computer readable instructions, data structures, program modules or other data, etc. in a modulated data signal such as a carrier wave or other transport mechanism, and Includes all information delivery media. The term modulated data signal means a signal in which one or more of the characteristics of the signal is set or changed so as to encode information in 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 implementing various aspects of the disclosure is shown including a computer 1102 , the computer 1102 including a processing unit 1104 , a system memory 1106 , and a system bus 1108 . do. A system bus 1108 couples system components, including but not limited to system memory 1106 , to the processing device 1104 . The processing device 1104 may be any of a variety of commercially available processors. Dual processor and other multiprocessor architectures may also be used as processing unit 1104 .

The system bus 1108 may be any of several types of bus structures that may further interconnect 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, EEPROM, etc., the BIOS is the basic input/output system (BIOS) 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) - this 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 (eg, for reading from or writing to removable diskette 1118), and an optical disk drive 1120 (eg, a CD-ROM) for reading from, or writing to, disk 1122, or other high capacity optical media, such as DVD. The hard disk drive 1114 , the magnetic disk drive 1116 , and the optical disk drive 1120 are connected to the system bus 1108 by the hard disk drive interface 1124 , the magnetic disk drive interface 1126 , and the 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. 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 media above refers to HDDs, removable magnetic disks, and removable optical media such as CDs or DVDs, those skilled in the art will use zip drives, magnetic cassettes, flash memory cards, cartridges, etc. It will be appreciated that other tangible computer-readable media such as etc. 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 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 in various commercially available operating systems or combinations of operating systems.

A user may enter commands and information into the computer 1102 via one or more wired/wireless input devices, for example, a pointing device such as a keyboard 1138 and 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 connected to the processing unit 1104 through an input device interface 1142 that is often connected to the system bus 1108, parallel ports, IEEE 1394 serial ports, game ports, USB ports, IR interfaces, It may be connected by other interfaces, etc.

A monitor 1144 or other type of display device is also coupled to the system bus 1108 via an interface, such as a video adapter 1146 . In addition to the monitor 1144, the computer typically includes 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 workstations, computing device computers, routers, personal computers, portable computers, microprocessor-based entertainment devices, peer devices, or other common network nodes, and are typically connected to computer 1102 . Although it includes many or all of the components described for it, only memory storage device 1150 is shown for simplicity. The logical connections shown include wired/wireless connections to a local area network (LAN) 1152 and/or a larger network, eg, a wide area network (WAN) 1154 . Such 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, for example, the Internet.

When used in a LAN networking environment, the computer 1102 is coupled to the local network 1152 through a wired and/or wireless communication network interface or adapter 1156 . Adapter 1156 may facilitate wired or wireless communication to LAN 1152 , which LAN 1152 also includes a wireless access point installed therein for communicating with wireless adapter 1156 . When used in a WAN networking environment, the computer 1102 may include a modem 1158, be connected to a communication computing device on the WAN 1154, or establish communications over the WAN 1154, such as over the Internet. have other means. A modem 1158 , which may be internal or external and a wired or wireless device, is coupled to the system bus 1108 via a 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 other means of establishing a communication link between the computers may be used.

The computer 1102 may be associated with any wireless device or object that is deployed and operates in wireless communication, for example, a printer, scanner, desktop and/or portable computer, portable data assistant (PDA), communication satellite, wireless detectable tag. It operates to communicate with any device or place, and phone. This includes at least Wi-Fi and Bluetooth wireless technologies. Accordingly, the communication may be a predefined structure as in a conventional network or may simply be an ad hoc communication between at least two devices.

Wi-Fi (Wireless Fidelity) makes it possible to connect to the Internet, etc. without a wired connection. Wi-Fi is a wireless technology such as cell phones that allows these devices, eg, computers, to transmit and receive data indoors and outdoors, ie anywhere within range 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 may operate in unlicensed 2.4 and 5 GHz radio bands, for example at 11 Mbps (802.11a) or 54 Mbps (802.11b) data rates, or in products that include both bands (dual band). .

One of ordinary skill in the art of this disclosure will understand that information and signals may be represented using any of a variety of different technologies and techniques. For example, data, instructions, instructions, information, signals, bits, symbols, and chips that may be referenced in the above description are voltages, currents, electromagnetic waves, magnetic fields or particles, optical field particles or particles, or any combination thereof.

Those of ordinary skill in the art of the present disclosure will recognize that the various illustrative logical blocks, modules, processors, means, circuits, and algorithm steps described in connection with the embodiments disclosed herein include electronic hardware, (convenience For this purpose, 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 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 upon the particular application and design constraints imposed on the overall system. A person skilled in the art of the present disclosure may implement the described functionality in various ways for each specific application, but such implementation decisions should not be interpreted as a departure from the scope of the present disclosure.

The various embodiments presented herein may be implemented as methods, apparatus, or articles 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 (eg, hard disks, floppy disks, magnetic strips, etc.), optical disks (eg, CDs, DVDs, etc.), smart cards, and flash drives. memory devices (eg, EEPROMs, cards, sticks, key drives, etc.). Also, 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 presented processes is an example of exemplary approaches. Based on design priorities, it is to be understood that the specific order or hierarchy of steps in the processes may be rearranged within the scope of the present disclosure. The appended method claims present elements of the various steps in a sample order, but are not meant to be limited to the specific 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 readily apparent to those skilled in the art, and the generic principles defined herein may be applied to other embodiments without departing from the scope of the present disclosure. Thus, the present disclosure is not intended to be limited to the embodiments presented herein, but is to be construed in the widest scope consistent with the principles and novel features presented herein.

Claims (10)

In the video data transmission method performed by a computing device including at least one processor,
obtaining a user preference including a preference weight for each of two or more quality indicators indicating video quality; and
determining a data transmission amount per unit time for the integrated quality reflecting the acquired user preference to reach a target quality through a reinforcement learning decision model;
including,
The integrated quality is

is determined by
remind

is a benefit vector among the components constituting the video quality vector, and

is a penalty vector among the components constituting the video quality vector,

is a partial weight vector including preference weights corresponding to the profit vector among at least one preference weight included in the preference weight vector;

is a partial weight vector including preference weights corresponding to the penalty vector among at least one preference weight included in the preference weight vector;

is an exponential weight for expressing a non-linear preference for the profit vector, and

represents the sum of the elements of the weight vector corresponding to the profit vector among the components of the preference weight vector,

represents the sum of the elements of the weight vector corresponding to the penalty vector among the constituent components of the preference weight vector,
How to transfer video data.
The method of claim 1,
The two or more quality indicators include
A bitrate value, a framerate value, a rebuffering time value indicating the time required to fill the buffer, and the difference between the time the video data was sent from the server and the time the client received the video data At least one of the delay time values representing
How to transfer video data.
delete The method of claim 1,
The reinforcement learning decision model includes an artificial neural network layer including at least one node, and the learning method of the reinforcement learning decision model includes:
obtaining, by the reinforcement learning decision model, state information related to video quality from an environment;
obtaining, by the reinforcement learning decision model, a user preference including a preference weight for each of two or more quality indicators;
determining, by the reinforcement learning decision model, a data transmission amount per unit time based on the state information and the user preference; and
obtaining, by the reinforcement learning decision model, a reward from the environment;
containing,
How to transfer video data.
The method of claim 1,
The reinforcement learning decision model is
Determining, based on a Pareto optimal solution, the amount of data transmission per unit time for the integrated quality reflecting the obtained user preference to reach a target quality,
How to transfer video data.
The method of claim 1,
When a preference weight value for at least one of the preference weight values for each of the two or more quality indicators included in the user preference is not obtained,
obtaining user preference distribution data based on a user preference set including at least one previously input user preference;
further comprising,
How to transfer video data.
7. The method of claim 6,
The user preference distribution data is
Obtained based on the user preference set through a semi-supervised learning (Semi-Supervised Learning) prediction model,
How to transfer video data.
7. The method of claim 6,
After dividing the user preference set into N predetermined preference subsets based on the obtained user preference distribution data, a first user preference for a first preference subset and a second user preference for a second preference subset calculating;
calculating a data transmission amount per first candidate unit time so that the integrated quality reflecting the first user preference reaches a target quality through a reinforcement learning decision model;
calculating a data transmission amount per second candidate unit time so that the integrated quality reflecting the second user preference reaches a target quality through a reinforcement learning decision model; and
determining a final data transmission amount per unit time based on the data transmission amount per first candidate unit time and the data transmission amount per second candidate unit time;
further comprising,
How to transfer video data.
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 transmitting moving picture data, the operations comprising:
obtaining a user preference including a preference weight for each of two or more quality indicators indicating video quality; and
determining a data transmission amount per unit time so that the integrated quality reflecting the user preference obtained through the reinforcement learning decision model reaches a target quality;
including,
The integrated quality is

is determined by
remind

is a benefit vector among the components constituting the video quality vector, and

is a penalty vector among the components constituting the video quality vector,

is a partial weight vector including preference weights corresponding to the profit vector among at least one preference weight included in the preference weight vector;

is a partial weight vector including preference weights corresponding to the penalty vector among at least one preference weight included in the preference weight vector;

is an exponential weight for expressing a non-linear preference for the profit vector, and

represents the sum of the elements of the weight vector corresponding to the profit vector among the components of the preference weight vector,

represents the sum of the elements of the weight vector corresponding to the penalty vector among the constituent components of the preference weight vector,
A computer program stored on a computer-readable storage medium.
A device for transmitting video data, comprising:
one or more processors;
Memory; and
network department;
including, and
The one or more processors,
Obtaining user preference including preference weights for each of two or more quality indicators indicating video quality, and
Determining the amount of data transmission per unit time for the integrated quality reflecting the acquired user preference through the reinforcement learning decision model to reach the target quality,
The integrated quality is

is determined by
remind

is a benefit vector among the components constituting the video quality vector, and

is a penalty vector among the components constituting the video quality vector,

is a partial weight vector including preference weights corresponding to the profit vector among at least one preference weight included in the preference weight vector;

is a partial weight vector including preference weights corresponding to the penalty vector among at least one preference weight included in the preference weight vector;

is an exponential weight for expressing a non-linear preference for the profit vector, and

represents the sum of the elements of the weight vector corresponding to the profit vector among the components of the preference weight vector,

represents the sum of the elements of the weight vector corresponding to the penalty vector among the constituent components of the preference weight vector,
A device for transmitting video data.

Images