JP7557161B2 - COMMUNICATION SYSTEM, COMMUNICATION METHOD, AND PROGRAM - Google Patents

Description

The present invention relates to a communication system, a communication method and a program.

In recent years, attempts have been made to apply machine learning and deep learning techniques to digitally modulated signals obtained at the receiver of a digital coherent optical transceiver in order to estimate the state of communication systems, such as monitoring the transmission characteristics of optical fiber communications and detecting anomalies.

For example, Non-Patent Document 1 proposes the use of a neural network that uses as input parameters a constellation visualized by a two-dimensional plot of real and imaginary axes of a digitally modulated signal obtained at a receiver of a digital coherent optical transceiver. More specifically, a method is proposed for using such a neural network to identify distortion of the constellation caused by state changes due to bending or pressure applied to an optical fiber.

There are also attempts to use neural networks that have learned the shape of constellations to monitor transmission characteristics. For example, Non-Patent Document 2 shows that by having a neural network learn feature quantities calculated from constellations, it is possible to estimate quantities that represent unknown transmission characteristics such as chromatic dispersion and optical signal-to-noise ratio.

Here, an example of neural network learning and estimation aimed at estimating optical signal-to-noise ratio (OSNR) and polarization mode dispersion (PMD) at the receiver of a digital coherent optical transceiver is explained using Figure 17.

In such neural network training, a constellation with known OSNR and PMD is input to the neural network. In such training, a process of tuning the weights of the neural network based on the output result of the neural network and the OSNR and PMD in the input constellation is repeated. As a result, the neural network learns the regularity of each label. A label means a pair of OSNR and PMD.

When an unknown, unlabeled constellation is input to a trained neural network, the neural network outputs the most appropriate label based on past patterns.

Changes in the shape of the constellation caused by changes in the condition of the optical fiber, such as bending or dirt on the optical fiber, or deterioration of the transmitter's laser, are differences that are too small to be distinguished by human vision. However, by utilizing neural networks, it may be possible to detect changes in the signal shape caused by changes in the condition of the optical fiber, and so research is being conducted as described above.

T. Tanaka, S. Kuwabara, H. Nishizawa, T. Inui, S. Kobayashi, A. Hirano, “Field demonstration of real-time optical network diagnosis using deep neural network and telemetry”, OFC2019, Tu2E.5. J. A. Jargon, X. Wu, H. Y. Choi, Y. Chung, and A. E. Willner, “Optical performance monitoring of QPSK data channels by use of neural networks trained with parameters derived from asynchronous constellation diagrams,” Optics Express vol. 18, no. 5 , March 2010.

In general, when using neural networks to classify images, the more distinct the features of the images used for training, the higher the accuracy of classification. However, changes in constellation shape due to changes in the state of the optical fiber are delicate. For this reason, in situations where there are many labels to be identified, such as when it is necessary to estimate not only whether the optical fiber is bent but also its bending diameter, or when it is necessary to identify end-face dirt in addition to bending, the accuracy may decrease when learning only from changes in constellation shape.

Therefore, more data is needed to improve the estimation accuracy. However, obtaining sufficient data can require a lot of effort. As a result, it can take a lot of effort to estimate the state of a communication system.

In view of the above circumstances, the present invention aims to provide a technology that reduces the effort required to estimate the state of a communication system.

One aspect of the present invention is a communication system comprising: a basic constellation image data acquisition unit that acquires basic constellation image data, which is image data of a constellation image of an input optical signal, based on the input optical signal; a difference acquisition unit that acquires the difference between the basic constellation image data and each of image data of one or more pre-prepared constellation images; and a training data generation unit that generates data including the difference as training data to be used for learning a mathematical model that estimates the state of a communication system to be estimated based on the difference.

One aspect of the present invention is a communication system comprising: a signal receiving unit that receives an optical signal transmitted by an optical transmitter of a communication system to be estimated, the optical signal being transmitted by the optical transmitter; a basic constellation image data acquisition unit that acquires basic constellation image data, which is image data of an image of a constellation of the optical signal, based on an input optical signal; a difference acquisition unit that acquires the difference between the basic constellation image data and each of image data of one or more pre-prepared constellation images; a training data generation unit that generates data including the difference as training data used for learning a mathematical model that estimates the state of the communication system to be estimated based on the difference; and an estimation unit that estimates the state of the object to be estimated based on the optical signal received by the signal receiving unit using the learned mathematical model obtained by a learning device that includes: a signal receiving unit that receives the optical signal transmitted by the optical transmitter of the communication system to be estimated, the optical signal being transmitted by the optical transmitter;

One aspect of the present invention is a communication method having a basic constellation image data acquisition step of acquiring basic constellation image data, which is image data of a constellation image of an input optical signal, based on the input optical signal; a difference acquisition step of acquiring the difference between the basic constellation image data and each of image data of one or more pre-prepared constellation images; and a training data generation step of generating data including the difference as training data to be used for learning a mathematical model that estimates the state of a communication system to be estimated based on the difference.

One aspect of the present invention is a communication method having a signal receiving step of receiving an optical signal transmitted by an optical transmitter of a communication system to be estimated, the optical signal being transmitted by the optical transmitter, based on the input optical signal; a basic constellation image data acquisition unit that acquires basic constellation image data, which is image data of an image of a constellation of the optical signal, based on the input optical signal; a difference acquisition unit that acquires the difference between the basic constellation image data and each of image data of one or more pre-prepared constellation images; a training data generation unit that generates data including the difference as training data used for learning a mathematical model that estimates the state of the communication system to be estimated based on the difference; and a learning unit that learns the mathematical model using the training data generated by the training data generation unit, to estimate the state of the communication system to be estimated based on the optical signal received in the signal receiving step, using the learned mathematical model obtained by a learning device that includes:

One aspect of the present invention is a program for causing a computer to function as the above-mentioned communication device.

The present invention makes it possible to reduce the effort required to estimate the state of a communication system.

FIG. 2 is a diagram illustrating an example of a hardware configuration of an estimation apparatus according to an embodiment. FIG. 2 is a diagram illustrating an example of the configuration of a control unit included in the estimation device of the embodiment. FIG. 2 is a diagram illustrating an example of a hardware configuration of a learning apparatus according to an embodiment. 5 is an explanatory diagram illustrating a difference data acquisition process executed by a control unit of the learning device in the embodiment. FIG. FIG. 2 is a diagram showing an example of a hardware configuration of a characteristic change device according to an embodiment. FIG. 2 is a diagram showing an example of the configuration of a control unit included in the learning device according to the embodiment. 4 is a flowchart showing an example of a flow of a process executed by the estimation device of the embodiment. 1 is a first flowchart illustrating an example of a flow of a process executed by the learning device of the embodiment. 11 is a second flowchart illustrating an example of the flow of a process executed by the learning device of the embodiment. FIG. 1 is a diagram showing an example of a communication system according to an embodiment. 1 is a first flowchart showing an example of a flow of processing executed by an integrated learning device in an embodiment. 11 is a second flowchart showing an example of the flow of processing executed by the integrated learning device in the embodiment. 1 is a first flowchart showing an example of a flow of a process executed by a communication system in an embodiment. 11 is a second flowchart showing an example of the flow of a process executed by the communication system in the embodiment. FIG. 1 is a first diagram showing an example of a configuration of a communication system according to an embodiment. FIG. 2 is a second diagram showing an example of the configuration of a communication system according to an embodiment. FIG.

(Embodiment)
1 is a diagram illustrating an example of a hardware configuration of an estimation device 1 according to an embodiment. The estimation device 1 receives an optical signal propagating in a communication system to be estimated, and estimates the state of the communication system to be estimated based on the received optical signal. The state of the communication system to be estimated is, for example, the state of an optical fiber.

The estimation device 1 has a control unit 11 including a processor 91 such as a CPU and a memory 92 connected by a bus, and executes a program. By executing the program, the estimation device 1 functions as a device including the control unit 11, a communication unit 12, a memory unit 13, a signal receiving unit 14, an output unit 15, and an optical branching unit 16.

More specifically, the processor 91 reads out a program stored in the storage unit 13 and stores the read out program in the memory 92. The processor 91 executes the program stored in the memory 92, whereby the estimation device 1 functions as a device including a control unit 11, a communication unit 12, a storage unit 13, a signal receiving unit 14, an output unit 15 and an optical branching unit 16.

The control unit 11 controls the operation of various functional units equipped in the estimation device 1. The control unit 11 records various information in the memory unit 13, for example. The control unit 11 executes a constellation image acquisition process, for example, on an optical signal received by the signal receiving unit 14. The constellation image acquisition process is a process for acquiring image data of a constellation image of the optical signal to be processed. Hereinafter, the image of the constellation is referred to as a constellation image. Furthermore, below, the image data of the constellation image is referred to as constellation image data.

The control unit 11 executes, for example, a difference data acquisition process. The difference data acquisition process is a process for acquiring difference data. The difference data is information indicating the difference between the constellation image data obtained by the constellation image acquisition process and the reference constellation image data. The reference constellation image data is image data that has been stored in advance in a predetermined storage device, and is image data of a constellation image. The predetermined storage device in the difference data acquisition process executed by the control unit 11 is, for example, the memory unit 13.

The difference information obtained by executing the difference acquisition process is, for example, information indicating the difference between the constellation image data obtained by the constellation image acquisition process and the reference constellation image data previously stored in the storage unit 13. More specifically, the information indicating the difference is, for example, information indicating the difference between a tensor representing the constellation image data and a tensor representing the reference constellation image data.

The control unit 11 executes a previously obtained trained estimation model on, for example, the acquired difference data. The trained estimation model is a mathematical model that estimates the state of the communication system to be estimated based on the difference data, and is a mathematical model obtained by learning. Therefore, the control unit 11 estimates the state of the communication system to be estimated by executing the trained estimation model on the difference data.

A mathematical model is a set including one or more processes whose execution conditions and order (hereinafter referred to as "execution rules") are predetermined. Learning means updating the mathematical model using machine learning methods. Updating the mathematical model means appropriately adjusting the parameter values in the mathematical model. Executing a mathematical model means executing each process included in the mathematical model in accordance with the execution rules.

The mathematical model is updated through learning until a predetermined termination condition for learning (hereinafter referred to as the "learning termination condition") is satisfied. The learning termination condition is, for example, a condition that a predetermined number of learning rounds have been performed.

The communication unit 12 includes a communication interface for connecting the estimation device 1 to an external device. The communication unit 12 communicates with the external device via a wired or wireless connection. The external device is communicatively connected to, for example, a computer operated by a user, and accepts user operations on the estimation device 1.

The memory unit 13 is configured using a computer-readable storage media device such as a magnetic hard disk device or a semiconductor storage device. The memory unit 13 stores various information related to the estimation device 1. The memory unit 13 stores, for example, a pre-trained estimation model. The memory unit 13 stores, for example, one or more reference constellation image data in advance. The signal receiving unit 14 receives an optical signal.

The output unit 15 outputs various information. The output unit 15 includes a display device such as a CRT (Cathode Ray Tube) display, a liquid crystal display, or an organic EL (Electro-Luminescence) display. The output unit 15 may be configured as an interface that connects these display devices to the estimation device 1. The output unit 15 outputs, for example, the results of execution of the learned estimation model by the control unit 11 in a predetermined output method such as display.

The optical branching unit 16 branches the optical signal. The optical branching unit 16 is, for example, an optical coupler. One of the optical signals branched by the optical branching unit 16 enters the signal receiving unit 14, and the other is emitted from the estimation device 1 and propagates toward a specified destination. Note that the estimation device 1 does not necessarily need to include the optical branching unit 16. In such a case, the optical signal that enters the estimation device 1 is not emitted from the estimation device 1.

2 is a diagram showing an example of the configuration of the control unit 11 provided in the estimation device 1 of the embodiment. The control unit 11 includes a constellation image data acquisition unit 111, a difference acquisition unit 112, an estimation unit 113, and an output control unit 114.

The constellation image data acquisition unit 111 executes a constellation image acquisition process on the optical signal received by the signal receiving unit 14. As a result, the constellation image data acquisition unit 111 acquires constellation image data of the optical signal received by the signal receiving unit 14.

The difference acquisition unit 112 executes a difference data acquisition process on the constellation image data acquired by the constellation image data acquisition unit 111. The difference acquisition unit 112 acquires difference data by executing a difference data acquisition process on the constellation image data acquired by the constellation image data acquisition unit 111.

The estimation unit 113 executes the learned estimation model on the difference data acquired by the difference acquisition unit 112. The estimation unit 113 estimates the state of the communication system to be estimated by executing the learned estimation model on the difference data.

The output control unit 114 controls the operation of the output unit 15. The output control unit 114 controls the operation of the output unit 15 to cause the output unit 15 to output the result of the estimation by the estimation unit 113.

<About the trained estimation model>
As described above, the trained estimation model is a mathematical model obtained by learning. More specifically, the trained estimation model is a mathematical model in which the estimation model is updated until the learning termination condition is satisfied. The estimation model is a mathematical model that estimates the state of the communication system to be estimated based on the difference data. Therefore, the trained estimation model is an estimation model at the point in time when the learning termination condition is satisfied.

The estimation model may be any mathematical model expressed as long as it can be updated by learning. The estimation model is expressed, for example, as a neural network. The parameters of the neural network are suitably adjusted based on the value of the objective function (i.e., the loss). The parameters of the neural network are the parameters of the mathematical model that it represents.

Here, we will explain the device that learns the estimation model. Figure 3 is a diagram showing an example of the hardware configuration of the learning device 2 of the embodiment. The learning device 2 is a device that obtains a learned estimation model by updating the estimation model through learning.

The learning device 2 has a control unit 21 including a processor 93 such as a CPU and a memory 94 connected by a bus, and executes a program. By executing the program, the learning device 2 functions as a device including the control unit 21, communication unit 22, memory unit 23, signal receiving unit 24, and output unit 25.

More specifically, the processor 93 reads out the program stored in the storage unit 23 and stores the read out program in the memory 94. The processor 93 executes the program stored in the memory 94, whereby the learning device 2 functions as a device including a control unit 21, a communication unit 22, a storage unit 23, a signal receiving unit 24 and an output unit 25.

The control unit 21 controls the operation of various functional units equipped in the learning device 2. The control unit 21 records, for example, various information in the memory unit 23. The control unit 21 executes a constellation image acquisition process, for example, on an optical signal received by the signal receiving unit 24. Hereinafter, the constellation image data obtained by the execution of the constellation image acquisition process by the control unit 21 is referred to as basic constellation image data.

The control unit 21, for example, records the obtained basic constellation image data in the storage unit 23. The control unit 21 reads out the basic constellation image data stored in the storage unit 23 at the timing when a predetermined condition (hereinafter referred to as the "reading condition") regarding the reading out of the basic constellation image data from the storage unit 23 is satisfied. The reading condition is, for example, a condition that the storage unit 23 has stored a predetermined number or more of basic constellation image data, each of which has a predetermined amount of information or more.

In this way, the basic constellation image data is stored in the memory unit 23 until the readout conditions are met, and one or more basic constellation image data are stored in the memory unit 23.

The control unit 21 executes a difference data acquisition process for each basic constellation image data that has been read out. The reference constellation image data in the difference data acquisition process executed by the control unit 21 is already stored in, for example, the memory unit 23.

Here, the difference data acquisition process executed by the control unit 21 will be explained using the diagram. Figure 4 is an explanatory diagram explaining the difference data acquisition process executed by the control unit 21 of the learning device 2 in the embodiment.

In Fig. 4, "basic constellation data group" means a set of basic constellation image data. Images represented by the basic constellation image data can be classified into a finite number of sets based on a predetermined criterion. "Label1" and "Label2" in Fig. 4 are identifiers of sets to which images of each basic constellation image data classified based on a predetermined criterion belong. In Fig. 4, each letter S having a subscript number and a superscript number, such as "S ₁ ⁰ ", indicates one basic constellation image data.

Set G1 means a set of reference constellation image data. In Figure 4, "differential constellation data group" means a set of difference data. As shown in Figure 4, in the difference data acquisition process performed by the control unit 21, the differences between one or more basic constellation image data and one or more reference constellation image data are respectively acquired.

In this way, the control unit 21 obtains difference data for each pair of basic constellation image data and reference constellation image data. Therefore, the control unit 21 generates difference data in a number equal to the product of the number of basic constellation image data and the number of reference constellation image data.

As shown in FIG. 4, at least a portion of the basic constellation image data may be identical to the reference constellation image data.

Returning to the explanation of FIG. 3, the control unit 21 executes an estimation model on the acquired difference data, for example. The control unit 21 updates the estimation model based on the result of executing the estimation model, for example.

The communication unit 22 includes a communication interface for connecting the learning device 2 to an external device. The communication unit 22 communicates with the external device via a wired or wireless connection. The external device is communicatively connected to, for example, a computer operated by a user, and accepts user operations on the learning device 2.

The memory unit 23 is configured using a computer-readable storage media device such as a magnetic hard disk device or a semiconductor storage device. The memory unit 23 stores various information related to the learning device 2. The memory unit 23 stores, for example, an estimation model in advance. The memory unit 23 stores, for example, one or more reference constellation image data in advance. The signal receiving unit 24 receives an optical signal.

The output unit 25 outputs various information. The output unit 25 is configured to include a display device such as a CRT display, a liquid crystal display, or an organic EL display. The output unit 25 may be configured as an interface that connects these display devices to the learning device 2. The output unit 25 outputs, for example, the results of execution of the estimation model by the control unit 21 in a predetermined output method such as display.

<Optical signal received by signal receiving unit 24>
Here, the optical signal received by the signal receiving unit 24 will be described. The optical signal received by the signal receiving unit 24 is an optical signal that is the result of a characteristic modification process being performed on an optical signal propagating within the communication system to be estimated. The characteristic modification process is a process that changes a target characteristic of the optical signal. The target characteristic is a predetermined characteristic of the optical signal. The characteristic may be, for example, chromatic dispersion, polarization mode dispersion, or signal-to-noise ratio. The target characteristic is, for example, the optical signal-to-noise ratio (OSNR) and polarization mode dispersion.

The change in the target characteristic is not the same for all optical signals. It is preferable that the change in the target characteristic is as close to random as possible. The change in the target characteristic may be a change that satisfies conditions specified by the user. Therefore, the learning device 2 can update the estimation model using multiple types of basic constellation image data.

Information indicating the content of the characteristic change due to the characteristic change process (hereinafter referred to as "change content information") is used to train the estimation model. Therefore, the communication unit 22 communicates with the device that executes the characteristic change process and acquires the change content information. Hereinafter, the device that executes the characteristic change process is referred to as the characteristic change device.

Figure 5 is a diagram showing an example of the hardware configuration of the characteristic change device 3 in an embodiment. The characteristic change device 3 has a control unit 31 having a processor 95 such as a CPU and a memory 96 connected by a bus, and executes a program. By executing the program, the characteristic change device 3 functions as a device having the control unit 31, communication unit 32, memory unit 33, optical characteristic change unit 34, and optical spectrum analyzer 35.

More specifically, the processor 95 reads out a program stored in the storage unit 33 and stores the read out program in the memory 96. The processor 95 executes the program stored in the memory 96, whereby the characteristic change device 3 functions as a device including a control unit 31, a communication unit 32, a storage unit 33, an optical characteristic change unit 34, and an optical spectrum analyzer 35.

The control unit 31 controls the operation of various functional units included in the characteristic change device 3. The control unit 31 controls, for example, the operation of the optical characteristic change unit 34. The control unit 31 controls, for example, the operation of the optical spectrum analyzer 35 to obtain the results of the analysis by the optical spectrogram analyzer. The control unit 31 records, for example, various information in the memory unit 33.

The communication unit 32 includes a communication interface for connecting the characteristic change device 3 to an external device. The communication unit 32 communicates with the external device via a wired or wireless connection. The external device is, for example, the learning device 2. The communication unit 32 transmits information indicating the content of the change made by the light characteristic change unit 34 to the learning device 2. The information indicating the content of the change made by the light characteristic change unit 34 is change content information. The communication unit 32 may also be communicatively connected to a computer operated by a user, and may accept user operations on the characteristic change device 3.

The storage unit 33 is configured using a computer-readable storage medium device such as a magnetic hard disk device or a semiconductor storage device. The storage unit 33 stores various information related to the characteristic change device 3. The storage unit 33 may store, for example, change content information.

An optical signal propagating within the communication system to be estimated is incident on the optical characteristics change unit 34. The optical characteristics change unit 34 executes a characteristic change process on the incident optical signal. In other words, the optical characteristics change unit 34 changes the characteristics of the incident optical signal.

In the example of FIG. 5, the optical characteristic change unit 34 includes a chromatic dispersion generator 36, a polarization mode dispersion generator 37, and an amplified spontaneous emission (ASE) light source 38. The chromatic dispersion generator 36 changes the chromatic dispersion. The polarization mode dispersion generator 37 changes the polarization mode dispersion (PMD). The ASE light source 38 adds noise to the signal.

The configuration of the optical characteristics change unit 34 in FIG. 5 is an example, and the optical characteristics change unit 34 may have any configuration as long as it can change the desired characteristics of the optical signal. Therefore, the optical characteristics change unit 34 may include, for example, only one of the chromatic dispersion generator 36, the polarization mode dispersion generator 37, or the ASE light source 38, or only two of them. The optical characteristics change unit 34 may also include, for example, a device that changes the characteristics of the optical signal other than the chromatic dispersion generator 36, the polarization mode dispersion generator 37, and the ASE light source 38.

The characteristic change device 3 outputs the optical signal converted by the optical characteristic change unit 34. The optical signal output by the characteristic change device 3 is incident on the learning device 2.

The control unit 31 acquires information indicating how the optical characteristics change unit 34 converted the optical signal (i.e., change content information) from the optical characteristics change unit 34. The control unit 31 transmits the acquired change content information to the learning device 2 via the communication unit 32. The learning device 2 receives the transmitted change content information via the communication unit 22.

The optical spectrum analyzer 35 analyzes the characteristics of the optical signal after the change by the optical characteristic change unit 34.

Figure 6 is a diagram showing an example of the configuration of the control unit 21 provided in the learning device 2 in an embodiment. The control unit 21 includes a change content information acquisition unit 210, a basic constellation image data acquisition unit 211, a memory control unit 212, a difference acquisition unit 213, a training data generation unit 214, and a mathematical model learning unit 215.

The change content information acquisition unit 210 acquires the change content information received by the communication unit 32.

The basic constellation image data acquisition unit 211 executes a constellation image acquisition process on the optical signal received by the signal receiving unit 24. The memory control unit 212 records the basic constellation image data obtained by executing the constellation image acquisition process in the memory unit 23.

When recording the basic constellation image data, the memory control unit 212 records the corresponding change content information in the memory unit 23 in association with the basic constellation image data. The corresponding change content information means change content information indicating the content of the characteristic change processing when an optical signal having the constellation indicated by the basic constellation image data as its characteristic is generated.

As an example, using optical signal P and optical signal Q obtained by executing a characteristic change process on optical signal P, basic constellation image data and change content information recorded in storage unit 23 in association with the basic constellation image data will be described. In such a case, the basic constellation image data recorded in storage unit 23 is constellation image data R of optical signal Q, and the change content information recorded in storage unit 23 is information R indicating the content of the change from optical signal P to optical signal Q. The constellation image data R and information R are recorded in storage unit 23 in association with each other.

The memory control unit 212 reads out the basic constellation image data stored in the memory unit 23 when the readout conditions are met.

The difference acquisition unit 213 executes a difference data acquisition process. By executing the difference data acquisition process, the difference acquisition unit 213 acquires, as difference data, each difference between each basic constellation image data read by the memory control unit 212 and each reference constellation image data previously stored in the memory unit 23.

The training data generation unit 214 associates each of the obtained difference data with change content information corresponding to the basic constellation image data used to obtain the difference data, and records the data in the storage unit 23. That is, the training data generation unit 214 records each piece of information including a pair of difference data and change content information corresponding to the difference data in the storage unit 23.

The change content information corresponding to the difference data means the change content information corresponding to the basic constellation image data used to obtain the difference data. Hereinafter, information containing a pair of difference data and change content information corresponding to the difference data is referred to as training data. The training data is used to learn the estimation model.

The processing of the training data generation unit 214 will be explained using a specific example. As an example, difference data C generated from basic constellation image data A and reference constellation image data B will be used. In such a case, the training data generation unit 214 associates the difference data C with change content information D corresponding to the basic constellation image data A and records it in the memory unit 23. Therefore, in this example, the training data is information including a pair of difference data C and change content information D.

In this way, the training data generation unit 214 generates training data using the difference data acquired by the difference acquisition unit 213 and the change content information corresponding to the basic constellation image data used to obtain the difference data.

The mathematical model learning unit 215 uses the training data to learn the estimation model. More specifically, the mathematical model learning unit 215 executes the estimation model on the difference data contained in the training data, and updates the estimation model so as to reduce the difference between the result of the execution and the change content information contained in the training data.

The mathematical model learning unit 215 includes a training data acquisition unit 216, a mathematical model execution unit 217, and an update unit 218. The training data acquisition unit 216 acquires training data. The mathematical model execution unit 217 executes an estimation model on difference data included in the training data acquired by the training data acquisition unit 216. The update unit 218 updates the estimation model so as to reduce the difference between the change content information associated with the difference data to be executed for the estimation model and the execution result of the estimation model.

Figure 7 is a flowchart showing an example of the flow of processing executed by the estimation device 1 of the embodiment. The signal receiving unit 14 receives an optical signal (step S101). Next, the constellation image data acquisition unit 111 executes a constellation image acquisition process on the optical signal received in step S101 (step S102). By executing step S102, the constellation image data acquisition unit 111 acquires constellation image data of the optical signal received in step S101.

Next, the difference acquisition unit 112 executes a difference data acquisition process (step S103). By executing the difference data acquisition process, the difference acquisition unit 112 acquires the difference between the constellation image data acquired in step S102 and the reference constellation image data as difference data. Next, the estimation unit 113 executes a trained estimation model (i.e., a trained estimation model) on the difference data acquired in step S103 (step S104). Note that the trained estimation model is, for example, obtained in advance from the learning device 2 and stored in the storage unit 13 in advance. After step S104, the output control unit 114 controls the operation of the output unit 15 to output the result of the execution of step S104 (step S105).

Figure 8 is a first flowchart showing an example of a process flow executed by the learning device 2 of the embodiment. More specifically, Figure 8 is a flowchart showing an example of a process flow in which the learning device 2 acquires training data. The signal receiving unit 24 receives the optical signal after the change by the characteristic change process (step S201). The characteristic change process is executed by, for example, the characteristic change device 3.

Next, the change content information acquisition unit 210 acquires change content information indicating the content of the characteristic change process that generated the optical signal in step S201 (step S202). Next, the basic constellation image data acquisition unit 211 executes a constellation image acquisition process on the optical signal received in step S201 (step S203). By executing the constellation image acquisition process on the optical signal received in step S201, the basic constellation image data acquisition unit 211 acquires basic constellation image data of the optical signal received in step S201.

Next, the memory control unit 212 records the basic constellation image data obtained in step S203 and the change content information obtained in step S202 in association with each other in the memory unit 23 (step S204). Next, the memory control unit 212 determines whether the readout condition is satisfied (step S205). If the readout condition is not satisfied (step S205: NO), the process returns to step S201. On the other hand, if the readout condition is satisfied (step S205: YES), the memory control unit 212 reads out each basic constellation image data stored in the memory unit 23 and each reference constellation image data stored in the memory unit 23 from the memory unit 23 (step S206).

Next, the difference acquisition unit 213 acquires each difference between each basic constellation image data acquired in step S206 and each reference constellation image data as difference data (step S207). Next, the training data generation unit 214 generates training data including a pair of each difference data acquired in step S207 and change content information corresponding to the basic constellation image data used to acquire the difference data (step S208). Specifically, the training data generation unit 214 records each difference data in the storage unit 23 in association with the change content information corresponding to the basic constellation image data used to acquire the difference data.

In this way, training data used to learn the estimation model is generated. Note that the processing from step S201 to step S203 does not necessarily have to follow the processing flow of FIG. 8, and may be performed in any order as long as the processing from step S201 to step S203 is performed before the processing of step S204. Note that the processing of step S203 needs to be performed after the processing of step S201 is performed.

Figure 9 is a second flowchart showing an example of the process flow executed by the learning device 2 of the embodiment. More specifically, Figure 9 is a flowchart showing an example of the process flow in which the learning device 2 learns an estimation model. The training data acquisition unit 216 acquires one of the training data stored in the memory unit 23 (step S301). Next, the mathematical model execution unit 217 executes the estimation model on the difference data included in the training data acquired by the training data acquisition unit 216 (step S302). By executing the estimation model, the mathematical model execution unit 217 estimates change content information corresponding to the difference data to be executed.

Next, the update unit 218 updates the estimation model so as to reduce the difference between the estimation result of the estimation model and the change content information included in the training data acquired by the training data acquisition unit 216 (step S303). The estimation result of the estimation model is specifically the change content information estimated in step S302. Next, the update unit 218 determines whether the learning end condition is satisfied (step S304).

If the learning end condition is met (step S304: YES), the process ends. The estimation model at the time when the learning end condition is met is the learned estimation model. On the other hand, if the learning end condition is not met (step S304: NO), the training data acquisition unit 216 acquires one unused training data from the training data stored in the memory unit 23 (step S305). After step S305, the process of step S302 is executed. Note that unused training data means training data that has not yet been used to learn the estimation model. More specifically, unused training data means training data that has not been the subject of execution of the estimation model.

The learning device 2 of the embodiment configured in this manner uses one or more pre-stored reference constellation image data to obtain the differences between one or more basic constellation image data and one or more reference constellation image data. Therefore, once one basic constellation image data is obtained, it is possible to generate training data equal to the number of reference constellation image data. In other words, the learning device 2 can reduce the effort required to prepare training data used to learn an estimation model.

The more training data there is, the more the accuracy of the estimation of the estimation model can be improved through learning. As described above, the trained estimation model is the estimation model at the point in time when the learning termination condition is met. Also, as described above, the estimation model is a mathematical model that estimates the state of the communication system to be estimated based on the difference data, and is updated using the difference data.

As described above, the difference data is information indicating the difference between the constellation image data of the optical signal and the reference constellation image data, and therefore the difference data is information indicating the difference in the constellation images. Therefore, the estimation model is a mathematical model that estimates the state of the communication system to be estimated based on the difference in the constellation images. For this reason, updating the estimation model improves the accuracy of identifying the constellation images by the estimation model.

Therefore, the learning device 2 of the embodiment configured in this manner can reduce the effort required to prepare training data used to learn the estimation model, and therefore the effort required to improve the accuracy of constellation image identification. Also, the learning device 2 can reduce the effort required to improve the accuracy of constellation image identification, and therefore the effort required to estimate the state of the communication system.

Furthermore, since the estimation device 1 of the embodiment configured in this manner uses the learned estimation model obtained by the learning device 2, it is possible to reduce the effort required from construction to operation of the communication system to be estimated, which is the effort required to estimate the state of the communication system.

<First application scene>
An example of an application scene of the estimation device 1, the learning device 2, and the characteristic modification device 3 will be described. Fig. 10 is a diagram showing an example of a communication system 100 in an embodiment. For the sake of simplicity of the following description, functional units having the same functions as those shown in Figs. 1 to 3, 5, and 6 are denoted by the same reference numerals as those shown in Figs. 1 to 3, 5, and 6, and description thereof will be omitted.

The communication system 100 is a communication system that achieves both communication and a reduction in the burden required for managing the device itself. The communication system 100 is, for example, a communication system that performs optical fiber communication. The communication system 100 comprises a transmitting optical node device 5, an integrated learning device 6, and an estimation device 1. The estimation device 1 provided in the communication system 100 comprises an optical branching unit 16. Therefore, the estimation device 1 provided in the communication system 100 outputs the incident optical signal toward the propagation destination.

The transmitting optical node device 5 transmits an optical signal. The integrated learning device 6 receives the optical signal and updates the estimation model through learning based on the received optical signal.

The sending optical node device 5 and the integrated learning device 6 are connected by an optical fiber 4, and an optical signal propagates from the sending optical node device 5 to the integrated learning device 6 through the optical fiber 4. The sending optical node device 5 and the estimation device 1 are connected by an optical fiber 4, and an optical signal propagates from the sending optical node device 5 to the estimation device 1 through the optical fiber 4. The optical fiber 4 is a transmission path (i.e., an optical fiber) through which the optical signal propagates.

The transmitting optical node device 5 includes an optical transmitter 51 and an optical branching unit 52. The optical transmitter 51 transmits an optical signal. The optical branching unit 52 branches the optical signal. The optical branching unit 52 is, for example, an optical coupler.

The integrated learning device 6 receives the branched optical signal branched by the optical branching unit 52. More specifically, the integrated learning device 6 receives one of the optical signals branched into multiple signals by the optical branching unit 52. The integrated learning device 6 includes a characteristic change device 3 and a learning device 2.

Note that the change in the characteristics of the optical signal by the characteristic change device 3 is not the same for all optical signals. It is preferable that the change in characteristics be as close to random as possible. The change in the characteristics of the optical signal by the characteristic change device 3 may be a change that satisfies conditions specified by the user. In this way, the characteristic change device 3 generates multiple types of optical signals.

The integrated learning device 6 updates the estimated model at a predetermined timing until the learning end condition is satisfied. Hereinafter, the predetermined timing is referred to as the learning timing. The learning timing is, for example, a periodic timing that occurs at a predetermined cycle. That is, the integrated learning device 6 performs a series of processes for updating the estimated model from the start of learning until the learning end condition is satisfied at each learning timing. Therefore, the estimated model that is updated at least at the second and subsequent learning timings is a learned estimated model for which the learning end condition was satisfied at the immediately preceding learning timing.

The estimation device 1 receives the branched optical signal branched by the optical branching unit 52. More specifically, the receiving optical node device 7 receives an optical signal that is one of the optical signals branched by the optical branching unit 52 and is different from the optical signal propagating to the integrated learning device 6.

An example of the process flow executed by the integrated learning device 6 will be described using Figures 11 and 12. Figure 11 is a first flowchart showing an example of the process flow executed by the integrated learning device 6 in an embodiment. Figure 12 is a second flowchart showing an example of the process flow executed by the integrated learning device 6 in an embodiment. Hereinafter, processes similar to those in Figures 7 to 9 will be given the same reference numerals as those in Figures 7 to 9 and descriptions thereof will be omitted.

The integrated learning device 6 determines whether it is time to learn (step S401). The determination of whether it is time to learn may be made by any of the functional units of the integrated learning device 6, for example, by the control unit 31 of the characteristic change device 3. The determination of whether it is time to learn may also be made by the control unit 21 of the learning device 2.

If it is not time to learn (step S401: NO), the process of step S101 is repeated. On the other hand, if it is time to learn (step S401: YES), the characteristic change device 3 receives the optical signal branched by the optical branching unit 52 (step S402). Next, the characteristic change device 3 changes the characteristics of the received optical signal (step S403). Next, the learning device 2 executes the processes of steps S201 to S204.

After the processing of step S204, the learning device 2 determines whether or not the read condition is satisfied (step S205a). If the read condition is not satisfied (step S205a: NO), the processing returns to step S402. On the other hand, if the read condition is satisfied (step S205a: YES), the processing of steps S206 to S208 is executed. After the processing of step S208, the processing of steps S301 to S305 is executed.

Figure 13 is a first flowchart showing an example of the flow of processing executed by the communication system 100 in an embodiment. More specifically, Figure 13 is a flowchart showing an example of the flow of processing executed by the communication system 100 at the learning timing.

The optical transmitter 51 transmits an optical signal (step S501). Next, the optical signal is split by the optical splitter 52 (step S502). One of the optical signals split by the optical splitter 52 propagates to the integrated learning device 6, and the other of the split optical signals propagates to the estimation device 1 (step S503). The integrated learning device 6 uses the propagated optical signal to update the estimation model until the learning end condition is satisfied (step S504). The integrated learning device 6 transmits the obtained trained estimation model to the estimation device 1 via the communication unit 22 (step S505). In step S504, for example, a series of processes described in FIG. 11 and FIG. 12 are executed.

Figure 14 is a second flowchart showing an example of the flow of processing executed by the communication system 100 in an embodiment. More specifically, Figure 14 is a flowchart showing an example of the flow of processing executed by the communication system 100 at a timing other than the learning timing.

The optical transmitter 51 transmits an optical signal (step S601). Next, the optical signal is split by the optical splitter 52 (step S602). One of the optical signals split by the optical splitter 52 propagates to the integrated learning device 6, and the other of the split optical signals propagates to the estimation device 1 (step S603). The signal receiver 14 receives the optical signal that has propagated to the estimation device 1 (step S604). The estimation device 1 estimates the state of the communication system 100 using the optical signal received (step S605). The process of step S605 is specifically the execution of the processes from step S102 to step S104. Next, the output control unit 114 controls the operation of the output unit 15 to cause the output unit 15 to output the result of the execution of step S605 (step S606).

<Relationship and Roles of the Integrated Learning Device 6, the Optical Splitter 52, and the Estimation Model>
Not only the communication system 100, but communication systems in general are expected to be used for a long period of time, and will deteriorate over time. For this reason, it is desirable to estimate the state of the communication system taking into account the effects of deterioration over time in order to manage the communication system. The state of the communication system is, for example, the state of the optical transmitter 51. For management of a system incorporating such changes over time, it is desirable to change the detection criteria and the rules for determining the state of the system in accordance with the changes over time in the system. For this reason, the communication system 100 includes an optical branching unit 52 and an integrated learning device 6.

As described above, the integrated learning device 6 changes the optical signal branched by the optical branching unit 52 and learns how the optical signal changes in the optical branching unit 52 at the learning timing. The change is caused by the characteristic change device 3.

The roles of the optical branching unit 52 and the characteristic change device 3 will be explained in order. In infrastructure facilities such as a communication system, it is difficult to install and remove devices after the start of a service. Specifically, the start of a service means the start of communication. For this reason, it is desirable for the integrated learning device 6 to be provided in the communication system 100 in a state where it does not need to be removed from the communication system 100.

Therefore, the communication system 100 includes an optical branching unit 52. Because the communication system 100 includes the optical branching unit 52, the optical signal transmitted by the optical transmitter 51 is branched by the optical branching unit 52, and the optical signal is transmitted to the integrated learning device 6 in addition to communication. As a result, in the communication system 100, the mathematical model can be updated in response to changes in the optical transmitter 51 over time in addition to communication.

Next, the role of the characteristic change device 3 will be explained. The change in the optical signal caused by the characteristic change device 3 is equivalent to simulating a change that may occur in the optical signal due to an event occurring within the communication system 100. Therefore, the accuracy of estimation by the estimation model is improved by learning using the optical signal after the characteristics have been changed by the characteristic change device 3.

Meanwhile, the target characteristics change mainly due to two factors with different time scales. One is the change over time in the state of the communication system 100. Hereinafter, the change over time in the state of the communication system 100 is referred to as the first factor. The other is the randomness contained in natural phenomena such as Heisenberg's uncertainty principle and noise. Hereinafter, the randomness contained in natural phenomena such as Heisenberg's uncertainty principle and noise is referred to as the second factor.

Changes due to the first factor have a longer time scale than changes due to the second factor. When managing a system that is expected to be used for a long period of time, such as communication system 100, these two types of changes must be taken into consideration in order to understand the state of communication system 100.

Since the integrated learning device 6 receives the optical signal branched from the optical branching unit 52, it is possible to update the estimation model even when the communication system 100 is performing communication. If the change in the first factor is to be ignored, it is sufficient to obtain once by machine learning a mathematical model that estimates the state of the communication system using the target characteristics of the input optical signal. The mathematical model obtained in this way is capable of estimating the state of the communication system with high accuracy, provided that the change in the communication system due to the first factor can be ignored.

However, if there is a change in the communication system due to the first factor, unless learning is performed in accordance with the changes in the communication system over time, the accuracy of the estimation will decrease as the communication system changes over time.

As described above, the integrated learning device 6 of the communication system 100 can update the estimation model even when the communication system 100 is communicating. Therefore, the integrated learning device 6 can perform learning not just once, but multiple times in response to changes over time in the communication system 100. Therefore, the integrated learning device 6 can obtain a mathematical model that estimates the state of the communication system 100 under the condition that the first factor and the second factor exist.

More specifically, we will explain how learning is possible under the condition that the first factor and the second factor exist. As described above, the characteristic change device 3 changes the optical signal branched by the optical branching unit 52. In other words, the characteristic change device 3 generates an optical signal by changing the optical signal branched by the optical branching unit 52.

When the state of the communication system 100 changes over time, the optical signal branched by the optical branching unit 52 changes in a manner that correlates with the change. Therefore, the characteristic change device 3 generates an optical signal that includes information on the change in the state of the communication system 100 over time.

On the other hand, the characteristic change device 3 imparts simulated changes that may occur in the optical signal due to events occurring within the communication system 100 as described above. These simulated changes mimic the changes due to the second factor. In other words, the characteristic change device 3 imparts changes in the optical signal caused by the randomness contained in natural phenomena to the optical signal received by the integrated learning device 6.

In this way, the characteristic change device 3 generates an optical signal including information on the change caused by the first factor and information on the change caused by the second factor. Since the target characteristics of the optical signal generated by the characteristic change device 3 are used for learning as part of the training data, the integrated learning device 6 can perform learning under the condition that the first factor and the second factor are present.

Learning by the integrated learning device 6 is performed on the estimation model. Then, as described above, the target characteristics of the optical signal generated by the characteristic change device 3 are used for learning as part of the training data. Therefore, the estimation model is a mathematical model that estimates the difference between the optical signal to be estimated for the target characteristics and the optical signal at the optical branching unit 52, and this difference is information that indicates the estimated result of the state of the communication system 100. In other words, the estimation model is a mathematical model that estimates the state of the communication system 100 based on the target characteristics of the input optical signal.

As described above, the trained mathematical model obtained by the integrated learning device 6 is used by the estimation device 1 to estimate the state of the communication system 100. The trained estimation model is further updated at the learning timing, and a new trained estimation model is generated. Since the estimation device 1 uses the estimation model obtained by the integrated learning device 6, the trained estimation model used by the estimation device 1 is updated to the trained estimation model at the time the learning termination condition was satisfied each time the learning termination condition is satisfied. Therefore, the communication system 100 enables management of the communication system taking into account not only the second factor but also the first factor.

The communication system 100 configured in this manner includes an optical branching unit 52 that branches the optical signal transmitted by the optical transmitter 51, and an integrated learning device 6 that updates the estimation model based on the optical signal branched by the optical branching unit 52. Therefore, the optical signal transmitted from the optical transmitter 51 is branched by the optical branching unit 52 and propagates to the integrated learning device 6 and the estimation device 1.

As a result, in the communication system 100, it is possible to realize the transmission of an optical signal from the optical transmitter 51 to the estimation device 1 while updating the mathematical model (i.e., the estimation model) that estimates the state of the communication system 100. Therefore, the communication system 100 can reduce the burden required for managing its own system.

Furthermore, since the communication system 100 includes the learning device 2, the effort required to estimate the state of the communication system can be reduced. Furthermore, the communication system 100 includes an estimation device 1 that estimates the state of the communication system 100 using the learned estimation model obtained by the learning device 2. Therefore, the communication system 100 can reduce the effort required to estimate the state of its own system.

<Second application scene>
Another example of an application scene of the estimation device 1, the learning device 2, and the characteristic change device 3 will be described. The estimation device 1, the learning device 2, and the characteristic change device 3 may be applied to a communication system that does not include an optical splitter 52. In such a case, the learning device 2 and the characteristic change device 3 are installed in the communication system when learning the estimation model, but are removed from the communication system at other times. Hereinafter, such a communication system that does not include an optical splitter 52 is referred to as a communication system 100a.

To simplify the following explanation, parts having the same functions as those described in Figures 1 to 3, 5, 6 and 10 will be given the same symbols as those in Figures 1 to 3, 5, 6 and 10, and their explanations will be omitted.

Figure 15 is a first diagram showing an example of the configuration of a communication system 100a in an embodiment. More specifically, Figure 15 is a diagram showing an example of the configuration of the communication system 100a at a timing other than when an estimation model is being learned. The communication system 100a includes an optical transmitter 51 and an estimation device 1. The estimation device 1 included in the communication system 100a includes an optical branching unit 16. Therefore, the estimation device 1 included in the communication system 100a outputs the incident optical signal toward the destination.

Figure 16 is a second diagram showing an example of the configuration of the communication system 100a in an embodiment. More specifically, Figure 16 is a diagram showing an example of the configuration of the communication system 100a when learning an estimation model. The communication system 100a includes an optical transmitter 51 and an integrated learning device 6.

In this way, since the communication system 100a includes the learning device 2, it is possible to reduce the effort required to estimate the state of the own system. Furthermore, the communication system 100a includes an estimation device 1 that estimates the state of the communication system 100a using the learned estimation model obtained by the learning device 2. Therefore, the communication system 100a can reduce the effort required to estimate the state of the own system.

(Modification)
It is not necessary for the communication system 100 to include only one optical transmitter 51 and one receiving-side optical node device 7. The communication system 100 may include a plurality of both the optical transmitters 51 and the receiving-side optical node devices 7.

The optical branching unit 52 does not necessarily have to be provided in the transmitting optical node device 5, and may be implemented in a housing different from that of the transmitting optical node device 5 and the optical branching unit 52.

Note that the estimation device 1, the learning device 2, and the characteristic change device 3 do not necessarily need to be mounted in different housings, and all three may be mounted in the same housing. Note that only two of the estimation device 1, the learning device 2, and the characteristic change device 3 may be mounted in the same housing, rather than all three.

The change content information is information indicating the content of the characteristic change due to the characteristic change process, in other words, information indicating how the optical signal transmitted by the optical transmitter 51 was changed.

Each of the estimation device 1, the learning device 2, and the characteristic change device 3 may be implemented using a plurality of information processing devices communicatively connected via a network. All or part of each function of the estimation device 1, the learning device 2, and the characteristic change device 3 may be realized using hardware such as an ASIC (Application Specific Integrated Circuit), a PLD (Programmable Logic Device), or an FPGA (Field Programmable Gate Array). The program may be recorded on a computer-readable recording medium. Examples of computer-readable recording media include portable media such as flexible disks, optical magnetic disks, ROMs, and CD-ROMs, and storage devices such as hard disks built into computer systems. The program may be transmitted via a telecommunications line.

Although an embodiment of the present invention has been described in detail above with reference to the drawings, the specific configuration is not limited to this embodiment and also includes designs that do not deviate from the gist of the present invention.

100, 100a...communication system, 1...estimation device, 11...control unit, 12...communication unit, 13...storage unit, 14...signal receiving unit, 15...output unit, 16...optical branching unit, 111...constellation image data acquisition unit, 112...difference acquisition unit, 113...estimation unit, 114...output control unit, 2...learning device, 21...control unit, 22...communication unit, 23...storage unit, 24...signal receiving unit, 25...output unit, 3...characteristic change device, 31...control unit, 32...communication unit, 33...storage unit, 34...optical characteristic change unit, 35...optical spectrum analyzer, 36...wavelength dispersion generator, 37...polarization mode dispersion generator, 38...ASE light source, 21...control unit, 210...change content information acquisition unit, 211: Basic constellation image data acquisition unit, 212: Memory control unit, 213: Difference acquisition unit, 214: Training data generation unit, 215: Mathematical model learning unit, 216: Training data acquisition unit, 217: Mathematical model execution unit, 218: Update unit, Transmitting side optical node device, 51: Optical transmitter, 52: Optical branching unit, 6: Integrated learning device, 4: Optical fiber, 91, 93, 95: Processor, 92, 94, 96: Memory

Claims (8)

a basic constellation image data acquisition unit that acquires basic constellation image data, which is image data of an image of a constellation of the optical signal, based on the input optical signal;
a difference acquisition unit that acquires a difference between the basic constellation image data and image data of one or more pre-prepared constellations;
a training data generating unit that generates data including the difference as training data used for learning a mathematical model that estimates a state of a communication system to be estimated based on the difference;
A communication system comprising: a learning unit that learns the mathematical model using the training data generated by the training data generation unit;
The communication system of claim 1 further comprising: The communication system includes an optical transmitter for transmitting an optical signal;
The constellation indicated by the basic constellation image data is a constellation of an optical signal in which the characteristics of the optical signal transmitted by the optical transmitter have been changed.
3. A communication system according to claim 1 or 2. a signal receiving unit for receiving an optical signal transmitted by an optical transmitter of a communication system to be estimated;
an estimation unit that estimates a state of the estimation target based on the optical signal received by the signal receiving unit, using the trained mathematical model obtained as a result of training of the mathematical model performed by the learning unit until a predetermined termination condition is satisfied;
The communication system of claim 2 further comprising: a signal receiving unit for receiving an optical signal transmitted by an optical transmitter of a communication system to be estimated;
an estimation unit that estimates a state of the estimation target based on the optical signal received by the signal receiving unit using the learned mathematical model obtained by a learning device including: a basic constellation image data acquisition unit that acquires basic constellation image data, which is image data of a constellation image of the optical signal based on an input optical signal; a difference acquisition unit that acquires a difference between the basic constellation image data and each of image data of one or a plurality of pre-prepared constellation images; a training data generation unit that generates data including the difference as training data used for learning a mathematical model that estimates a state of a communication system to be estimated based on the difference; and a learning unit that performs learning of the mathematical model using the training data generated by the training data generation unit;
A communication system comprising: a basic constellation image data acquisition step of acquiring basic constellation image data, which is image data of an image of a constellation of the optical signal, based on the input optical signal;
a difference obtaining step of obtaining a difference between the basic constellation image data and each of image data of one or more pre-prepared constellations;
a training data generating step of generating data including the difference as training data used for learning a mathematical model that estimates a state of the communication system to be estimated based on the difference;
A communication method comprising: a signal receiving step of receiving an optical signal transmitted by an optical transmitter of a communication system to be estimated, the optical signal being transmitted by the optical transmitter;
an estimation step of estimating a state of the estimation target based on the optical signal received in the signal receiving step, using the learned mathematical model obtained by a learning device including: a basic constellation image data acquisition unit that acquires basic constellation image data, which is image data of an image of a constellation of the optical signal, based on an input optical signal; a difference acquisition unit that acquires a difference between the basic constellation image data and each of image data of one or a plurality of pre-prepared constellation images; a training data generation unit that generates data including the difference as training data used for learning a mathematical model that estimates a state of a communication system to be estimated based on the difference; and a learning unit that learns the mathematical model using the training data generated by the training data generation unit;
A communication method comprising: A program for causing a computer to function as a communication system according to any one of claims 1 to 5.

Images