Sorting information generation device, signal sorting device, and sorting information generation method

Abstract

To adjust a ratio of the number of signals to be sorted to each device.SOLUTION: A sorting information generation device 1 comprises: a first setting unit 10 which sets distribution of observation data which is a collection of observed values, with respect to a mixture probability model which sets presence / absence of sorting of a signal with respect to each of a plurality of opposing devices 3 as the observed value of each cluster; a second setting unit 11 which sets a parameter of the mixture probability model so as to become the distribution of the set observation data; a learning unit 12 which sets the mixture probability model having the set parameter, as true sorting information related to the presence / absence of the sorting of the signal with respect to each of a plurality of devices, and performs adversarial learning of a generation model; and a presentation unit 14 which presents signal sorting information including a learned generator 121' constructed by the learning unit 12.SELECTED DRAWING: Figure 1

Description

[Technical Field]

[0001] The present invention relates to a distribution information generating device, a signal distribution device, and a distribution information generating method. [Background technology]

[0002] A round-robin load balancer has been known as a technology for distributing loads by distributing signals to multiple devices. For example, Patent Document 1 discloses a system that instructs an edge router to change the destination server to a backup server group according to the load status of an operational server group, and then uses a load balancer to distribute signals evenly to the backup server group of the changed destination.

[0003] The devices receiving the assigned signals may have different hardware resource capacities, such as the number of CPU cores. In such cases, it may be desirable to bias the number of signals assigned to each device depending on the hardware resource capacities. However, with the system disclosed in Patent Document 1, it was difficult to adjust the number of signals assigned to each device and assign signals to each device with the desired bias. [Prior art documents] [Patent documents]

[0004] [Patent Document 1] Japanese Patent Application Publication No. 2018-142848 Summary of the Invention [Problem to be solved by the invention]

[0005] As described above, with the conventional technology, it has been difficult to adjust the ratio of the number of signals allocated to each device.

[0006] The present invention has been made to solve the above-mentioned problems, and has as its object to adjust the ratio of the number of signals allocated to each device. [Means for solving the problem]

[0007] In order to solve the above-mentioned problems, the distribution information generation device of the present invention includes: a first setting unit configured to set a distribution of observation data, which is a set of observation values, for a mixed probability model in which the presence or absence of signal distribution for each of a plurality of devices is set as an observation value for each cluster; a second setting unit configured to set parameters of the mixed probability model so as to obtain the set distribution of observation data; a generator that uses the mixed probability model having the set parameters as true distribution information regarding the presence or absence of signal distribution for each of the plurality of devices and generates pseudo distribution information similar to the true distribution information; a learning unit configured to perform adversarial learning of a generative model having a classifier that distinguishes between the pseudo distribution information generated by the generator and the true distribution information; and a presentation unit configured to present signal distribution information including the trained generator constructed by the learning unit.

[0008] In addition, in the allocation information generation device according to the present invention, the mixed probability model may be a model that follows a mixed Bernoulli distribution, and the parameters of the mixed probability model may include a mixture ratio that represents the probability that each cluster will generate the observed value, and a probability that the observed value of each cluster will result in the signal being allocated.

[0009] In the distribution information generating device according to the present invention, the first setting unit may set a probability of occurrence of the observed value for each cluster.

[0010] In addition, in the distribution information generating device of the present invention, the plurality of devices may be functional nodes in a control plane of a core network that conforms to a predetermined communication standard, and the signal may include a control signal used in the control plane.

[0011] In order to solve the above-mentioned problems, the signal distribution device of the present invention comprises a generation unit configured to generate the pseudo distribution information using the learned generator included in the signal distribution information presented by the above-mentioned distribution information generation device, an identification unit configured to identify a device to which the signal is to be distributed from among the plurality of devices based on the pseudo distribution information generated by the generation unit, and a signal distribution unit configured to distribute the signal to the identified device.

[0012] In order to solve the above-mentioned problems, the communication control method of the present invention includes a first setting step of setting a distribution of observation data, which is a set of observation values, for a mixed probability model in which the presence or absence of signal allocation to each of a plurality of devices is set as an observation value for each cluster; a second setting step of setting parameters for the mixed probability model so that the set distribution of observation data is the set distribution of observation data; a learning step of performing adversarial learning of a generative model using the mixed probability model having the set parameters as true allocation information regarding the presence or absence of signal allocation to each of the plurality of devices, a generator that generates pseudo-allocation information similar to the true allocation information, and a classifier that distinguishes between the pseudo-allocation information generated by the generator and the true allocation information; and a presentation step of presenting signal allocation information including the trained generator constructed in the learning step.

[0013] In the communication control method according to the present invention, the first setting step may set a probability that the observed value occurs for each cluster. [Effects of the Invention]

[0014] According to the present invention, a mixture probability model having set parameters is used as true allocation information regarding whether or not a signal is allocated to each of a plurality of devices, and adversarial learning of a generative model having a generator that generates pseudo-allocation information similar to the true allocation information and a classifier that distinguishes between the pseudo-allocation information generated by the generator and the true allocation information is performed. As a result, the ratio of the number of signals allocated to each device can be adjusted. [Brief explanation of the drawings]

[0015] [Figure 1] FIG. 1 is a block diagram showing the configuration of a signal control system including a distribution information generating device and a signal distribution device according to an embodiment of the present invention. [Figure 2] FIG. 2 is a diagram for explaining an outline of the signal control system according to this embodiment. [Figure 3] FIG. 3 is a diagram for explaining an outline of the signal control system according to this embodiment. [Figure 4] FIG. 4 is a diagram for explaining the learning unit included in the allocation information generating device according to this embodiment. [Figure 5] FIG. 5 is a diagram for explaining the learning unit included in the allocation information generating device according to this embodiment. [Figure 6] FIG. 6 is a diagram for explaining the learning unit included in the allocation information generating device according to this embodiment. [Figure 7] FIG. 7 is a block diagram showing the hardware configuration of the distribution information generating device according to this embodiment. [Figure 8] FIG. 8 is a block diagram showing the hardware configuration of the signal distribution device according to this embodiment. [Figure 9] FIG. 9 is a sequence diagram showing the operation of the signal control system according to this embodiment. [Figure 10] FIG. 10 is a flowchart showing the operation of the distribution information generating device according to this embodiment. DETAILED DESCRIPTION OF THE INVENTION

[0016] DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Preferred embodiments of the present invention will now be described in detail with reference to FIGS.

[0017] [Configuration of signal control system] First, with reference to FIGS. 1 and 2, an outline of a signal control system including a distribution information generating device 1 and a signal distribution device 2 according to an embodiment of the present invention will be described.

[0018] The signal control system is provided in, for example, a 5G mobile communication system, and distributes and transmits signals such as control signals for performing connection management, authentication, mobility management, etc. in a core network. The distribution information generating device 1 and the signal distribution device 2 are connected via a network NW such as a WAN or the Internet. The distribution information generating device 1 generates signal distribution information and presents it to the signal distribution device 2 via the network NW.

[0019] The signal distribution device 2 is realized by, for example, an AMF (Access and Mobility Management Function) provided in the core network. The AMF is a function node of the control plane that manages the registration of user terminals and wireless connections. As shown in Fig. 2, the signal distribution device 2 is communicably connected to multiple opposite devices (devices) 3, and distributes signals to the multiple opposite devices 3 based on signal distribution information presented by the distribution information generating device 1.

[0020] The opposite device 3 is realized by, for example, a UDM (Unified Data Management) provided in the core network. The UDM is a functional node that manages subscriber information in the core network and manages the mobility of user terminals. Specifically, the AMF distributes a location registration request signal transmitted from the user terminal via a base station and transmits it to multiple UDMs. In this embodiment, D opposite devices 3 (D is an integer equal to or greater than 2) are provided.

[0021] The opposing device 3 can be realized by a computer equipped with a processor, main memory device, communication interface, auxiliary memory device, and input / output (I / O), and a program that controls these hardware resources. Multiple opposing devices 3 can each be equipped with hardware resources of different capacities. Each opposing device 3 is identified by an opposing device ID (#1, #2, . . . , #D).

[0022] The signal control system according to this embodiment employs a mixed probability model in which the bias in the number of signals that the signal distribution device 2 distributes to each of the multiple opposite devices 3 is modeled using a mixed Bernoulli distribution. The signal control system uses the mixed probability model, in which parameters are set to reflect the adjusted ratio of the number of signals to be distributed to each of the multiple opposite devices 3, as true data for adversarial learning to train a generator 121 that generates pseudo data similar to the true data. The signal control system also presents the trained generator 121' to the signal distribution device 2 as signal distribution information indicating to which opposite device 3 the signal distribution device 2 should distribute the signal. The signal distribution device 2 then distributes the signal to the opposite device 3 identified by the signal distribution information.

[0023] [Function block of the distribution information generation device] Next, functional blocks of the distribution information generating device 1 according to this embodiment will be described with reference to the block diagram of Fig. 1. As shown in Fig. 1, the distribution information generating device 1 includes a first setting unit 10, a second setting unit 11, a learning unit 12, a first storage unit 13, and a presentation unit 14.

[0024] The first setting unit 10 sets a distribution of observed data, which is a set of observed values, for a mixed probability model in which the observed value of each cluster is whether or not a signal is allocated to each of a plurality of opposed devices 3. In this embodiment, a model following a mixed Bernoulli distribution is adopted as the mixed probability model. The observed value is a value that represents, as binary data, whether or not a signal is allocated to each opposed device 3.

[0025] The mixed Bernoulli distribution employed in this embodiment is a model for clustering a set of N pieces of observed data formed by vectors of length D whose elements are 0 and 1. In this embodiment, the bias with which signals are allocated to a plurality of opposite devices 3 is considered as a plurality of clusters, and the observed data is a set of observed values ​​that express with 0 or 1 whether or not to allocate a signal to each of the plurality of opposite devices 3. In addition, the observed values ​​are generated from a plurality of different clusters having a binary distribution.

[0026] Here, we have N binary vector observations x1,x2,…,x N The set of observed data X is given by the parameters μ1,μ2,…,μ K The set of observed data X follows a mixed Bernoulli distribution with a parameter data set M. That is, it is assumed that the observed data X is generated from multiple clusters, and each cluster is associated with a parameter μ that indicates the probability that the observed value corresponds to the signal classification.

[0027] The observation data X and the parameter data set M are expressed as the following equations (1) and (2), respectively (T is a transpose symbol).

number

[0028] nth binary vector observation x n and the kth parameter vector μ k Each of these is further expressed as a vector having D elements, as shown in the following equations (3) and (4).

number

[0029] Observation x n element x of n As described above, each of (i) (i=1, 2, ..., D) is a binary variable that takes the value of 0 or 1. In this embodiment, the element x n (i) is defined as follows:

number

[0030] The probability of allocating a signal and the probability of not allocating a signal are expressed by the following equations (6) and (7), respectively.

number

[0031] From the above equations (6) and (7), the observed value x of the observed data X n When cluster k is assigned to observation x n is the parameter μ of cluster k k It follows the Bernoulli distribution of the following equation (8):

number

[0032] Here, the parameter data set π of the occurrence frequency of each cluster (K ​​clusters) is defined as follows: The parameter π is the observed value x n is a parameter that indicates how much of the data is distributed to each cluster. k is the observed value x of the observed data X n indicates the mixture ratio occurring in cluster k.

number

[0033] Also, the observed value x n When cluster k is assigned to observation x n is μ k It follows D Bernoulli distributions with parameters. Furthermore, each observation x n are independently generated by the mixed Bernoulli distribution of the following equation (10).

number

[0034] The parameter μ, which indicates the probability of the Bernoulli distribution in each cluster, is defined by the definition of the mixed Bernoulli distribution in equation (10). k and the cluster mixing ratio π k and from all clusters, observation x n The probability of occurrence of the above equation (10) is a mixed probability model in this embodiment, and represents the bias in the number of signals distributed to each of 1 to D opposite devices 3.

[0035] The first setting unit 10 sets the observed value x n The number and frequency of elements that take "distribute signals: 1" among the clusters 1 to K can be set according to a desired signal distribution pattern. Also, the first setting unit 10 sets the observed values ​​x that take "distribute signals: 1" across all of the clusters 1 to K. n Therefore, the first setting unit 10 can set the distribution of the observed data X that reflects the desired signal distribution pattern for each of a plurality of different clusters. n By setting the occurrence probability of the above, it becomes possible to set the signal distribution ratio for each of the D opposite devices 3.

[0036] FIG. 3 is a diagram for explaining the setting of the distribution of the observation data X by the first setting unit 10. Clusters 1 to K shown in FIG. 3 represent the entire signal distribution pattern for D opposed devices 3. In this embodiment, clusters 1 to K correspond to opposed devices 3 with IDs "#1" to "#D" respectively (K=D). "1 to D" constituting each cluster represent the observed values ​​x indicating whether or not to distribute signals to 1 to D opposed devices 3. n In the example of Fig. 3, for each piece of data 1 to D in each cluster, there are set opposite devices 3 to which signals are distributed (black dots in the figure) and opposite devices 3 to which signals are not distributed (data other than black dots in the figure).

[0037] The vertically aligned opposite device IDs between the clusters shown in Figure 3 correspond to each other, and in each of clusters 1 to K, 1 to D opposite devices 3 are set as signal distribution destinations. For example, in cluster 1, signals are distributed to the opposite device 3 with ID "#1", and in cluster 2, signals are distributed to the opposite device 3 with ID "#2". Furthermore, cluster K indicates that signals are distributed to the opposite device 3 with ID "#D".

[0038] The first setting unit 10 can set clusters in which signals are distributed at equal ratios among 1 to D opposite devices 3 in order to equalize the load. Alternatively, the first setting unit 10 can set clusters so that the desired signal distribution ratios are achieved for 1 to D opposite devices 3 according to the resource capacity of each device. For example, among D observations belonging to cluster 1, the element x n The value of (i) can be set so that signal allocation is enabled. Also, the proportion of observed data X that cluster 1 corresponding to the opposite device 3 with ID "#1" occupies can be set higher than that of other clusters (i.e., other opposite devices 3). In this way, in addition to the pattern of observed values ​​for each cluster, it is possible to set patterns such as the proportion of observed data X that each cluster occupies across multiple clusters.

[0039] Returning to FIG. 1, the second setting unit 11 sets the parameters of the mixed probability model so as to obtain the distribution of the observed data X set by the first setting unit 10. More specifically, the second setting unit 11 calculates the parameters μ when the values ​​of the observed data X set by the first setting unit 10 are given to the mixed probability model of the above formula (10). k , π k Find the value of .

[0040] For example, as shown in FIG. 3, when the first setting unit 10 sets a high ratio of the number of signals to be distributed to the opposite device 3 with ID "#1" belonging to cluster 1, the second setting unit 11 increases the value of the parameter π1 in the above equation (9) and also increases the value of each parameter μ kBy equalizing the clusters, the clusters from cluster 1 to x in the mixed probability model of equation (10) are n can be set to increase the probability of occurrence.

[0041] In this way, the second setting unit 11 determines the parameter μ k , π k By adjusting and setting n We can adjust the probability of occurrence of observation x from each cluster. n The mixed probability model in which the probability of occurrence of is set in advance is used as true data during adversarial learning by the learning unit 12 described below.

[0042] The learning unit 12 learns the parameter μ set by the second setting unit 11. k , π k The mixed probability model having the above formula is used as true allocation information regarding whether or not a signal is allocated to each of a plurality of opposite devices 3, and adversarial learning of a generative model is performed. The generative model includes a generator 121 that generates pseudo allocation information similar to the true allocation information, and a classifier 122 that distinguishes between the pseudo allocation information generated by the generator 121 and the true allocation information. The true allocation information is information that reflects a desired signal allocation ratio. The learning unit 12 can construct a trained generator 121' (trained generators 121'_1, . . . , 121'_K) for each of 1 to K clusters of the mixed probability model. The trained generators 121'_1, . . . , 121'_K indicate allocation information that represents the ratio of the number of signals to be allocated to opposite devices 3 with IDs "#1" to "#D" corresponding to clusters 1 to K, respectively.

[0043] As shown in FIG. 4, the learning unit 12 adversarially trains a GAN (Generative Adversarial Network) having a generator 121 and a classifier 122. In this embodiment, it is possible to provide as many pairs of generators 121 and classifiers 122 as there are clusters (K pairs). Alternatively, as shown in FIG. 4, it is possible to provide one classifier 122 for K generators 121_1, . . . , 121_K. Through learning by the learning unit 12, K trained generators 121′ (trained generators 121′_1, . . . , 121′_K) are constructed.

[0044] 5 and 6 are diagrams schematically illustrating the neural network configuration of the generator 121 and the classifier 122 of the GAN used by the learning unit 12. As shown in FIG. 5, the generator 121 is configured as a neural network having an input layer, a hidden layer, and an output layer. Note that the K generators 121_1, . . . , 121_K each have the same network structure, and will be collectively referred to as the generator 121 below. The generator 121 is a model that generates pseudo-sorting information from random noise. For example, m Gaussian noise vectors are randomly sampled and input to the input node of the generator 121 (z1 to z m ).

[0045] The generator 121 performs a product-sum operation on the input and weight parameters and performs threshold processing using an activation function to output an output G(z). The output G(z) from the generator 121 is calculated based on the parameter μ set by the second setting unit 11. k , π k The data is similar to the observed data X obtained by a mixture probability model having the set parameter μ k , π k The observed values ​​x of each cluster obtained by a mixture probability model with n Each output G(z1) to G(z D ) is the observation x of each cluster n As the neural network that configures the generator 121, a CNN or a ResNet can be used.

[0046] 6 is configured as a neural network having an input layer, a hidden layer, and an output layer. In the example of FIG. 6, the parameter μ k , π k The observed values ​​x of each cluster are obtained by a mixture probability model with n is given.

[0047] The classifier 122 performs a product-sum operation on the input and weight parameters and threshold processing using an activation function, and outputs a binary output of 1 or 0. When the classifier 122 correctly identifies the training data related to the input true classification information as true classification information, it outputs an output y=1. On the other hand, when the classifier 122 correctly identifies the training data related to the input pseudo classification information as pseudo classification information, it outputs an output y=0. In this way, the classifier 122 is a model that distinguishes the model distribution generated by the generator 121 from the data distribution of the training data, which is the true distribution. A CNN can be used as the neural network that constitutes the classifier 122.

[0048] FIG. 4 is a block diagram for explaining adversarial learning of GAN by the learning unit 12. The generator 121 of the GAN adopted by the learning unit 12 is represented by function G, and the classifier 122 is represented by function D. Furthermore, the true classification information is represented by x, the predicted value output by the classifier 122 is represented by y, and the correct label is represented by t. The correct label t is set to 1 for the true classification information and 0 for the pseudo classification information generated by the generator 121. At this time, the classifier 122 calculates the cross entropy E CE It can be expressed as:

[0049]

number

[0050] The first term in the brace of the above equation (11) represents t n lny n In this case, the predicted value y nis the correct label t of the true classification information. n = 1. On the other hand, the second term in the braces represents (1-t n )ln(1-y n ), the predicted value y n is the value of the correct label that identifies the pseudo-sorting information (1-t n ) = 0. In this way, the cross entropy E CE is the maximum value when the predicted value matches the correct label value.

[0051] Here, the generator 121 that constitutes the GAN has parameters w G ,θ G and the function G(w G ,θ G ) The classifier 122 also uses the parameter w D ,θ D and function D(w D ,θ D ) The cross entropy E in the above equation (11) CE The objective function E of the GAN including the generator 121 and the discriminator 122 based on the above can be expressed by the following equation (12).

number

[0052] The first term of the above equation (12) represents E D(x)=1 lnD(w D ,θ D ) is the expected value at which the classifier 122 classifies the true sorting information as the true sorting information. D(x)=0 ln(1-D(G(w G ,θ G ),w D ,θ D)) is an expected value at which the classifier 122 classifies the pseudo-sorting information generated by the generator 121 as pseudo-sorting information. In GAN learning, the generator 121 and the classifier 122 are trained in an adversarial manner by min-max optimization of the objective function E. Therefore, the generator 121 is trained so as to be able to generate pseudo-sorting information that will deceive the classifier 122, and the classifier 122 is trained so as to classify the pseudo-sorting information generated by the generator 121 as pseudo-sorting information.

[0053] In learning of the classifier 122, when true allocation information is given, the classifier 122 outputs an output close to y=1, thereby maximizing the first term of the objective function E in the above equation (12). On the other hand, when pseudo allocation information is given, the classifier 122 learns to output an output close to y=0, thereby maximizing the second term of the objective function E.

[0054] In the learning of the generator 121, D(G(w G ,θ G ),w D ,θ D ) (D(G(z)) in Figure 4) is close to 1. G ,θ G ) (G(z) in FIG. 4 ) to minimize the objective function E. The learning unit 12 uses a learning procedure that alternately updates the parameters of the generator 121 and the classifier 122. Details of the learning procedure of the generator 121 and the classifier 122 by the learning unit 12 will be described later.

[0055] The first storage unit 13 stores a trained generator 121′ in which the objective function E of the GAN has been optimized by the learning unit 12. More specifically, the first storage unit 13 stores K trained generators 121′ (trained generators 121′_1, . . . , 121′_K) corresponding to clusters 1 to K. The first storage unit 13 also stores a parameter μ k , π k The mixed probability model of the above equation (10) is stored.

[0056] The presentation unit 14 presents signal distribution information including the trained generator 121′ constructed by the learning unit 12. More specifically, the presentation unit 14 can transmit the signal distribution information to the signal distribution device 2 via the network NW. More specifically, the presentation unit 14 can transmit the signal distribution information including K trained generators 121′_1, . . . , 121′_K to the signal distribution device 2 via the network NW.

[0057] [Signal distribution device function block] Next, a description will be given of the functional blocks of the signal distribution device 2. As shown in Fig. 1, the signal distribution device 2 includes an acquisition unit 20, a second storage unit 21, a generation unit 22, an identification unit 23, and a signal distribution unit 24.

[0058] The acquisition unit 20 acquires the signal allocation information presented by the allocation information generation device 1. Specifically, the acquisition unit 20 can acquire the learned generators 121′ (learned generators 121′_1, . . . , 121′_K) of the signal allocation information stored in the second storage unit 21.

[0059] The second storage unit 21 stores the signal allocation information presented by the allocation information generating device 1. Specifically, the second storage unit 21 stores the learned generators 121′ (learned generators 121′_1, . . . , 121′_K).

[0060] The generation unit 22 generates pseudo allocation information using the trained generator 121′ included in the signal allocation information presented by the allocation information generation device 1. More specifically, the generation unit 22 generates pseudo allocation information using each of the trained generators 121′_1, . . . , 121′_K.

[0061] The pseudo-distribution information generated by the generating unit 22 is the observed value x nAs shown in the data set, the pseudo-allocation information generated by the learned generator 121′_1 corresponding to the cluster 1 and the opposite device 3 with ID “#1” is information that assigns a value of “distribute signal: 1” and a value of “do not distribute signal: 0” to each of the opposite devices 3 with ID “#1”, for example, permits the distribution of signals to the opposite device 3 with ID “#1”, and does not permit the distribution of signals to the opposite devices 3 with ID “#2” to “#D” (x n =[x n (1), x n (2), x n (3),···,x n (D)]=[1,0,0,···,0]).

[0062] The identifying unit 23 identifies the opposite device 3 to which the signal distribution device 2 distributes a signal, based on each of the K pieces of pseudo distribution information generated by the generating unit 22. Specifically, the identifying unit 23 identifies the element x n Among (i), the ID of the opposite device 3 having a value of "1" is identified.

[0063] The signal distribution unit 24 distributes signals to the identified opposite devices 3. For example, the signal distribution unit 24 registers all opposite devices 3 in a list and updates the index of the identified opposite device 3 based on each of the K pieces of pseudo distribution information. When a new signal is received, the signal distribution unit 24 selects the opposite device 3 indicated by the index, distributes the signal, and transmits it. When the next signal is received, the signal is similarly distributed to the opposite device 3 identified by the identification unit 23.

[0064] [Hardware configuration of the distribution information generation device] Next, an example of a hardware configuration for realizing the distribution information generating device 1 having the above-described functions will be described with reference to FIG.

[0065] 7, the distribution information generating device 1 can be realized by, for example, a computer including a processor 102, a main memory device 103, a communication interface 104, an auxiliary memory device 105, and an input / output (I / O) 106, which are connected via a bus 101, and a program for controlling these hardware resources. The distribution information generating device 1 further includes a display device 107.

[0066] The processor 102 is realized by a CPU, a GPU, an FPGA, an ASIC, or the like.

[0067] The main memory device 103 pre-stores programs for the processor 102 to perform various controls and calculations. The processor 102 and the main memory device 103 implement the functions of the allocation information generation device 1, such as the first setting unit 10, the second setting unit 11, the learning unit 12, and the presentation unit 14 shown in FIG.

[0068] The communication interface 104 is an interface circuit for connecting the distribution information generating device 1 to various external electronic devices via a network.

[0069] The auxiliary storage device 105 is composed of a readable / writable storage medium and a drive for reading and writing various information such as programs and data from and to the storage medium. The auxiliary storage device 105 can use a semiconductor memory such as a hard disk or flash memory as the storage medium.

[0070] The auxiliary storage device 105 has a program storage area for storing a signal control program that generates signal distribution information and presents it to the signal distribution device 2. The auxiliary storage device 105 also has a program storage area for storing a learning program for performing GAN adversarial learning executed by the distribution information generation device 1. The auxiliary storage device 105 also has a program storage area for storing a mixed probability model calculation program that calculates a mixed probability model related to a mixed Bernoulli distribution. The auxiliary storage device 105 realizes the first storage unit 13 described in FIG. 1. Furthermore, for example, the auxiliary storage device 105 may have a backup area for backing up the above-mentioned data, programs, etc.

[0071] The input / output I / O 106 is an input / output device that inputs signals from external devices and outputs signals to external devices.

[0072] The display device 107 is configured by an organic EL display, a liquid crystal display, etc. The display device 107 can display true sorting information on the screen.

[0073] [Hardware configuration of the signal distribution device] Next, an example of a hardware configuration for realizing the signal distribution device 2 having the above-described functions will be described with reference to FIG.

[0074] 8, the signal distribution device 2 can be realized by a computer including a processor 202, a main memory device 203, a communication interface 204, an auxiliary memory device 205, and an input / output (I / O) 206 connected via a bus 201, and a program for controlling these hardware resources, similar to the distribution information generation device 1. Furthermore, the signal distribution device 2 includes a display device 207.

[0075] The main memory device 203 pre-stores programs for the processor 202 to perform various controls and calculations. The processor 202, which is realized by a CPU, a GPU, an FPGA, an ASIC, or the like, and the main memory device 203 realize the functions of the signal distribution device 2, such as the acquisition unit 20, the generation unit 22, the identification unit 23, and the signal distribution unit 24 shown in FIG.

[0076] The communication interface 204 is an interface circuit for connecting the signal distribution device 2 to various external electronic devices via a network.

[0077] The auxiliary storage device 205 is composed of a readable / writable storage medium and a drive for reading and writing various information such as programs and data from and to the storage medium. The auxiliary storage device 205 can use a semiconductor memory such as a hard disk or flash memory as the storage medium.

[0078] The auxiliary storage device 205 has a program storage area for storing a generation program that performs calculations on the learned generator 121' based on the signal distribution information. The auxiliary storage device 105 also has a program storage area for storing a signal control program for distributing signals to the identified opposite device 3. The auxiliary storage device 205 realizes the second storage unit 21 described in FIG. 1. Furthermore, for example, the auxiliary storage device 205 may have a backup area for backing up the above-mentioned data, programs, etc.

[0079] The input / output I / O 206 is an input / output device that inputs signals from external devices and outputs signals to external devices.

[0080] The display device 207 is configured by an organic EL display, a liquid crystal display, or the like.

[0081] [Signal control system operation sequence] Next, the operation of the signal control system including the distribution information generating device 1 and the signal distribution device 2 having the above-described configuration will be described with reference to the sequence of FIG.

[0082] 9, first, the first setting unit 10 of the distribution information generating device 1 sets a distribution of the observation data X desired in the mixed probability model (step S1). Specifically, the first setting unit 10 sets the observed values ​​x of the observation data X, which specify the number of times and frequency of signal distribution to each of the 1 to D counterpart devices 3 for each cluster, so that the intended distribution distribution of signals is reflected. n (x n =[x n (1), x n (2), x n (3),···,x n In this way, in step S1, the observed value x n The values ​​of the observed data X can be set so that the probability of occurrence is equal or has a specific bias.

[0083] Next, the second setting unit 11 adjusts the parameter μ of the mixture probability model so as to obtain the distribution of the observation data X set in step S1. k , π k Specifically, the second setting unit 11 sets the observed value x for each cluster in the mixed probability model of the above formula (10). n When the observed data X is given with the probability of occurrence of k , π k The parameter μ is adjusted and determined. k , π k The mixed probability model in which the above is set is stored in the first storage unit 13.

[0084] Next, the learning unit 12 performs a learning process (step S3). Specifically, the learning unit 12 performs a learning process on the parameter μ k , π kAs true allocation information regarding whether or not a signal is allocated to each of D opposed devices 3, a mixed probability model having the above formula is used, and adversarial learning of a GAN having a generator 121 that generates pseudo allocation information similar to the true allocation information, and a classifier 122 that distinguishes between the pseudo allocation information generated by the generator 121 and the true allocation information is performed. Through the learning process, K trained generators 121'_1, . . . , 121'_K corresponding to clusters 1 to K, i.e., opposed devices 3 with IDs "#1" to "#D", respectively, are constructed and stored in the first storage unit 13. Details of the learning process in step S3 will be described later.

[0085] Next, the presenting unit 14 presents the signal distribution information including the trained generator 121′ constructed by the learning unit 12 in step S3 to the signal distribution device 2 via the network NW (step S4). Specifically, the presenting unit 14 presents K trained generators 121_1, . . . , 121′_K constructed for each cluster to the signal distribution device 2.

[0086] Subsequently, when signal distribution information is presented from the distribution information generating device 1, the signal distribution device 2 stores the learned generator 121′ (learned generators 121_1, . . . , 121′_K) including the signal distribution information in the second storage unit 21. Thereafter, the acquisition unit 20 of the signal distribution device 2 reads and acquires the learned generator 121′ (learned generators 121_1, . . . , 121′_K) from the second storage unit 21 (step S5).

[0087] Next, the generation unit 22 generates pseudo distribution information using the learned generators 121′ (learned generators 121_1, . . . , 121′_K) acquired in step S5 (step S6). Thereafter, the identification unit 23 identifies the opposite device 3 to which the signal distribution device 2 distributes the signal, based on the pseudo distribution information generated in step S6 (step S7). Next, the signal distribution unit 24 distributes the signal to the opposite device 3 identified in step S7 (step S8).

[0088] In this way, by distributing signals to the associated devices 3 specified in accordance with the signal distribution information, signals can be distributed to each associated device 3 at the intended distribution ratio of the number of signals.

[0089] Next, the learning process (step S3) by the learning unit 12 of the allocation information generating device 1 described in FIG. 9 will be described with reference to FIG. 10. In the example of the learning process shown in FIG. 10, the learning unit 12 repeatedly learns true allocation information for each of 1 to K clusters. First, the learning unit 12 inputs the true allocation information to the classifier 122 as training data 124, and adjusts the parameter w of the classifier 122 so that the true allocation information is distinguished from the true allocation information (y=1). D ,θ D is learned and updated (step S20).

[0090] As shown in the block diagram of the learning unit 12 in FIG. 4, the parameter μ set in step S2 in FIG. k , π k is true allocation information regarding whether or not a signal is allocated to each of the D opposite devices 3, and is used as training data 124 when training the classifier 122. In step S20, first, 1 to D observed values ​​x n is used as the training data 124.

[0091] In step S20, the learning unit 12 can cause the classifier 122 to learn the true allocation information using, for example, backpropagation. By step S20, the classifier 122 that can distinguish the true allocation information of cluster 1 from the true allocation information is constructed in advance.

[0092] Next, the learning unit 12 generates Gaussian noise and provides a random vector of the generated Gaussian noise as an input to the generator 121 (step S21). Subsequently, the generator 121 generates a random vector of the input z and the weight parameter w based on the provided Gaussian noise. G ,θ G, 121_K) (point a in FIG. 4), and generates one piece of pseudo allocation information G(z).

[0093] Next, the learning unit 12 performs learning of the classifier 122. The learning of the classifier 122 is performed by using the parameter w D ,θ D First, the learning unit 12 performs the learning process by fixing the parameter μ set in step S2 of FIG. k , π k The true classification information is input to the classifier 122 as training data 124. Then, the learning unit 12 adjusts the parameters w by backpropagation or the like so that the objective function E in the above equation (12) is maximized. D ,θ D (Step S23). The label of the training data 124 is set to 1 (true classification information). In step S23, first, the learning unit 12 updates the 1 to D observed values ​​x n The classifier 122 is trained using the above as true classification information.

[0094] Next, the learning unit 12 provides the pseudo-sorting information generated by the generator 121 in step S22 to the discriminator 122 as an input, and calculates the parameter w by backpropagation or the like so that the objective function E in the above equation (12) is maximized. D ,θ D (Step S24). That is, in steps S23 and S24, in order to maximize the objective function E in the above equation (12), the first term is updated as D(w D ,θ D )=1 is output, and the second term is D(G(w G ,θ G ),w D ,θ D)=0. In the training data 124, the label is set to 0 (pseudo allocation information). In step S24, the classifier 122 is trained using one pseudo allocation information G(z) generated by each of the K generators 121_1, . . . , 121_K and combined.

[0095] The learning of the classifier 122 in steps S23 and S24 corresponds to the dashed arrows in the block diagram of the learning unit 12 shown in FIG. 4 , which indicate that a classifier error is calculated in block 125 of the objective function E based on output 123 from the classifier 122, and then the error is back-propagated to the classifier 122.

[0096] Next, the learning unit 12 learns the generators 121 (generators 121_1, . . . , 121_K). The learning of the generators 121 is performed with the parameters of the discriminator 122 fixed. The learning unit 12 learns the generator 121 so that pseudo-sorting information is generated when random Gaussian noise is given to the generator 121. Specifically, the learning unit 12 learns the parameters w G ,θ G is updated (step S25).

[0097] The learning in step S25 corresponds to the flow of the dashed arrow indicating backpropagation of error to the generator 121 in the block diagram of the learning unit 12 in Fig. 4. That is, step S25 corresponds to the flow of the dashed arrow in which pseudo allocation information generated by the generator 121 in Fig. 4 is input to the discriminator 122, a generator error is calculated from its output 123 in the block 125 of the objective function E, and the error is further backpropagated to the generator 121.

[0098] Thereafter, learning of the discriminator 122 and the generator 121 (generators 121_1 to 121_K) from step S22 to step S25 is repeated until the value of the objective function E reaches a Nash equilibrium and converges (step S26: NO). On the other hand, if the value of the objective function E converges (step S26: YES), the processing from step S20 to step S26 is repeated using the remaining true allocation information from cluster 2 to cluster K out of all the true allocation information from cluster 1 to cluster K, in order, until learning of K generators 121_1, . . . , 121_K and the discriminator 122 is performed (step S27: NO).

[0099] Thereafter, when the generators 121 (generators 121_1 to 121_K) and the discriminator 122 are trained using the true allocation information of the remaining K-1 clusters from cluster 2 to cluster K (step S27: YES), the training unit 12 stores 1 to K trained generators 121'_1, ..., 121'_K in the first storage unit 13 (step S28). The trained generator 121' is constructed by the above-described processing from step S20 to step S28. After that, the processing proceeds to step S4 in FIG. 9.

[0100] As described above, according to the distribution information generating device 1 of this embodiment, the deviation in the number of signals distributed by the signal distribution device 2 to each of the plurality of opposite devices 3 is modeled by a mixed Bernoulli distribution, and the parameter μ is set so that the distribution of the observation data X representing the intended distribution is obtained. k , π k The mixed probability model with the set value is used as the true data in the adversarial learning of the GAN. Furthermore, since the trained generator 121′ constructed by the adversarial learning is presented to the signal distribution device 2 as signal distribution information, the ratio of the number of signals distributed to each opposing device 3 can be adjusted.

[0101] Furthermore, according to the distribution information generating device 1 of this embodiment, the parameter μ k , π kThe mixed probability model in which the above is set is used as true sorting information, which is true data in the adversarial learning of the GAN, and a generator 121 that generates pseudo sorting information similar to the true sorting information is trained. Therefore, by performing the adversarial learning of the GAN, it is possible to learn the latent variables of the mixed probability model that follows the mixed Bernoulli distribution.

[0102] Furthermore, according to the distribution information generating device 1 of this embodiment, the bias in the number of signals distributed to each of the multiple opposite devices 3 is used as the observed value of each cluster of the mixed Bernoulli distribution, so it is possible to specify the frequency and pattern of signal distribution to each opposite device 3 for each cluster. This makes it possible to set in detail the ratio of the number of signals distributed to each opposite device 3 according to the capacity of the hardware resources, thereby achieving more efficient signal control.

[0103] Furthermore, according to the distribution information generation device 1 according to this embodiment, the trained generator 121′ constructed by the adversarial learning of the GAN is notified as signal distribution information to the signal distribution device 2 via the network NW. Therefore, signal control can be performed remotely.

[0104] In the embodiment described above, the case where the signal distribution device 2 is an AMF and the opposite device 3 is a UDM has been exemplified. However, the signal distribution device 2 may be configured as, for example, a base station, and the opposite device 3 may be an AMF. Alternatively, the signal distribution device 2 may be a UPF (User Plane Function), and the opposite device 3 may be a cloud base.

[0105] In the described embodiment, the signal is exemplified as a control signal such as a location registration request signal for performing connection management, authentication, mobility management, etc. in the core network. However, the signal may be a data signal processed in the user plane.

[0106] The above describes embodiments of the distribution information generating device, signal distribution device, and distribution information generating method of the present invention, but the present invention is not limited to the described embodiments, and various modifications that a person skilled in the art can imagine can be made within the scope of the invention described in the claims. [Explanation of symbols]

[0107] 1...distribution information generation device, 2...signal distribution device, 10...first setting unit, 11...second setting unit, 12...learning unit, 13...first memory unit, 14...presentation unit, 20...acquisition unit, 21...second memory unit, 22...generation unit, 23...identification unit, 24...signal distribution unit, 101, 201...bus, 102, 202...processor, 103, 203...main memory unit, 104, 204...communication interface, 105, 205...auxiliary memory unit, 106, 206...input / output I / O, 107, 207...display device, 121...generator, 122...identifier, 123...output, 124...training data, 125...block of objective function E, NW...network.

Claims

1. a first setting unit configured to set a distribution of observation data, which is a set of observation values, for a mixed probability model in which the presence or absence of signal allocation to each of a plurality of devices is set as an observation value of each cluster; a second setting unit configured to set parameters of the mixed probability model so as to obtain a set distribution of the observation data; a learning unit configured to perform adversarial learning of a generative model including: a generator that generates pseudo-allocation information similar to true allocation information regarding whether or not the signals are allocated to each of the plurality of devices using the mixed probability model having the set parameters; and a classifier that distinguishes between the pseudo-allocation information generated by the generator and the true allocation information; a presentation unit configured to present signal distribution information including the trained generator constructed by the training unit; A distribution information generating device comprising:

2. 2. The distribution information generating device according to claim 1, the mixed probability model is a model that follows a mixed Bernoulli distribution, The parameters of the mixture probability model include a mixture ratio representing the probability that each cluster generates the observed value, and a probability that the observed value of each cluster corresponds to the signal classification. A distribution information generating device characterized by:

3. 2. The distribution information generating device according to claim 1, The first setting unit sets a probability of occurrence of the observed value for each cluster. A distribution information generating device characterized by:

4. 2. The distribution information generating device according to claim 1, the plurality of devices are function nodes in a control plane of a core network that conforms to a predetermined communication standard; The signals include control signals used in the control plane. A distribution information generating device characterized by:

5. a generation unit configured to generate the pseudo-allocation information using the trained generator included in the signal allocation information presented by the allocation information generation device according to any one of claims 1 to 4; an identification unit configured to identify a target device to which the signal is to be allocated from among the plurality of devices based on the pseudo allocation information generated by the generation unit; a signal distribution unit configured to distribute the signal to the identified device; A signal distribution device comprising:

6. a first setting step of setting a distribution of observed data, which is a set of observed values, for a mixed probability model in which the presence or absence of signal allocation to each of a plurality of devices is set as an observed value for each cluster; a second setting step of setting parameters of the mixed probability model so as to obtain the set distribution of the observation data; a learning step of performing adversarial learning of a generative model having a generator that generates pseudo-allocation information similar to true allocation information regarding whether or not the signals are allocated to each of the plurality of devices using the mixed probability model having the set parameters, and a classifier that distinguishes between the pseudo-allocation information generated by the generator and the true allocation information; a presentation step of presenting signal distribution information including the trained generator constructed in the learning step; A method for generating distribution information comprising the steps of:

7. 7. The method for generating distribution information according to claim 6, The first setting step sets a probability of occurrence of the observed value for each cluster. A method for generating distribution information.

Images