Optimization device, optimization method, and program

Abstract

To reduce the amount of calculation required to optimize the model to be optimized. [Solution] The optimization device according to the present disclosure includes a storage means for pre-storing a plurality of first input / output pairs, which are pairs of input and output signals of a machine learning model when an acoustic signal or vibration signal indicating a time-series change in sound or vibration observed by optical fiber sensing is input to the machine learning model as an input signal, and an evaluation means for optimizing the model to be optimized using the plurality of first input / output pairs.

Description

[Technical Field]

[0001] The present disclosure relates to an optimization device, an optimization method, and a program. [Background technology]

[0002] Optical fiber sensing, typified by Distributed Acoustic Sensing (DAS), is a technology that can observe sound and vibration at multiple points along an optical fiber.

[0003] In recent years, a technology has been proposed in which a machine learning model such as a deep learning model learns acoustic or vibration signals that indicate time-series changes in sound or vibration observed by optical fiber sensing. In this case, the learning target can be, for example, events occurring at points along the optical fiber or noise components contained in the acoustic or vibration signals.

[0004] Here, machine learning models such as deep learning models are nonlinear models trained on large amounts of data, and are characterized by having statistical knowledge obtained from large amounts of data.

[0005] For this reason, in recent years, a technique has been proposed in which statistical knowledge possessed by a machine learning model is incorporated into the optimization target model to optimize the optimization target model. The optimization target model is a model that is optimized using, for example, a mathematical optimization method or a machine learning model. For example, Patent Document 1 discloses a technique in which a neural process model, which is a deep learning model, is trained on the optimization target model. [Prior art documents] [Patent documents]

[0006] [Patent Document 1] International Publication No. 2022 / 250053 Summary of the Invention [Problem to be solved by the invention]

[0007] However, while machine learning models have the advantage of having statistical knowledge, they also require a large amount of calculations, which is why the optimization method disclosed in Patent Document 1 has the problem of increasing the amount of calculations required to optimize the model to be optimized.

[0008] In view of the above-mentioned problems, an object of the present disclosure is to provide an optimization device, an optimization method, and a program that can reduce the amount of calculation required to optimize a model to be optimized. [Means for solving the problem]

[0009] An optimization apparatus according to one aspect includes: a storage means for storing in advance a plurality of first input / output pairs, which are pairs of input and output signals of a machine learning model when an acoustic signal or a vibration signal indicating a time-series change in acoustics or vibration observed by optical fiber sensing is input as an input signal to the machine learning model; and an evaluation means for optimizing a model to be optimized using the plurality of first input / output pairs.

[0010] In one aspect, the optimization method comprises: An optimization method executed by an optimization device, comprising: storing in advance a plurality of first input / output pairs, which are pairs of input and output signals of a machine learning model when an acoustic signal or a vibration signal indicating a time-series change in acoustics or vibration observed by optical fiber sensing is input as an input signal to the machine learning model; optimizing a model to be optimized using the plurality of first input / output pairs.

[0011] In one aspect, the program comprises: On the computer, a step of storing in advance a plurality of first input / output pairs, which are pairs of input and output signals of a machine learning model when an acoustic signal or a vibration signal indicating a time-series change in acoustics or vibration observed by optical fiber sensing is input as an input signal to the machine learning model; optimizing a model to be optimized using the plurality of first input / output pairs; Execute the following. [Effects of the Invention]

[0012] According to the above-described aspects, it is possible to provide an optimization device, an optimization method, and a program that can reduce the amount of calculation required to optimize a model to be optimized. [Brief explanation of the drawings]

[0013] [Figure 1] FIG. 1 is a diagram illustrating the concept of an optimization device according to the present disclosure. [Figure 2] FIG. 1 is a block diagram illustrating a schematic configuration example of an optimization device according to the present disclosure. [Figure 3] FIG. 1 is a diagram illustrating a schematic operation example of an optimization device according to the present disclosure. [Figure 4] FIG. 4 is a flowchart illustrating an example of the operation flow of step S11 in FIG. 3. [Figure 5] 10A and 10B are diagrams illustrating the effects of the optimization device according to the present disclosure. [Figure 6] FIG. 1 is a diagram illustrating the concept of an optimization device according to the present disclosure. [Figure 7] FIG. 1 is a block diagram illustrating a schematic configuration example of an optimization device according to the present disclosure. [Figure 8] FIG. 1 is a diagram illustrating a schematic operation example of an optimization device according to the present disclosure. [Figure 9] 9 is a flowchart illustrating an example of the operation flow of step S21 in FIG. 8. FIG. [Figure 10] FIG. 1 is a block diagram illustrating a schematic configuration example of an optimization device according to the present disclosure. [Figure 11]FIG. 1 is a block diagram illustrating a schematic hardware configuration example of a computer that realizes an optimization device according to the present disclosure. DETAILED DESCRIPTION OF THE INVENTION

[0014] Hereinafter, embodiments of the present disclosure will be described with reference to the drawings. Note that the following description and drawings have been omitted and simplified as appropriate for clarity of explanation. In addition, in the following drawings, the same elements are given the same reference numerals, and duplicate explanations are omitted as necessary.

[0015] <First Embodiment> First, the concept of the first embodiment will be described. FIG. 1 is a diagram illustrating the concept of an optimization device 10 according to the present disclosure.

[0016] As shown in Figure 1, the optimization device 10 is a device that incorporates statistical knowledge possessed by a machine learning model such as a deep learning model (hereinafter, for convenience, referred to as a machine learning model M) into a model 20 to be optimized, and optimizes the model 20.

[0017] Here, the machine learning model M is a trained model that has been trained using as input signals acoustic signals or vibration signals that indicate time-series changes in acoustics or vibrations occurring at points along an optical fiber, as observed by optical fiber sensing, such as DAS.

[0018] For example, the machine learning model M is a trained model that uses an acoustic signal or a vibration signal as an input signal to learn the relationship between the acoustic signal or the vibration signal and text. In this case, the text is text such as an event occurring at a point along the optical fiber. In this case, the output signal of the machine learning model M is text or a pair of the acoustic signal or the vibration signal and the text.

[0019] Alternatively, the machine learning model M is a trained model that uses an acoustic signal or a vibration signal as an input signal to learn the noise components contained in the acoustic signal or the vibration signal. In this case, the output signal of the machine learning model M is a signal in which noise has been suppressed from the acoustic signal or the vibration signal.

[0020] For ease of explanation, the following description will be given assuming that the input signal of the machine learning model M is an acoustic signal that indicates the time series changes in the acoustics generated at points along the optical fiber, as observed by optical fiber sensing.

[0021] The optimization device 10 includes an input / output storage 11. The input / output storage 11 stores a plurality of input / output pairs (first input / output pairs) {(I k ,O k )} k=1~K Here, when the kth (k=1 to K) acoustic signal is input to the machine learning model M, the input / output pair of the machine learning model M is (I k ,O k )

[0022] The optimization device 10 calculates a plurality of input / output pairs {(I k ,O k )} k=1~K The model 20 is optimized while evaluating the likelihood, which indicates whether the output signal y of the model 20 when an input signal x, which is an acoustic signal, is input to the model 20, is plausible (close to the desired solution).

[0023] Next, the configuration of the first embodiment will be described. FIG. 2 is a block diagram showing a schematic configuration example of the optimization device 10 according to the present disclosure.

[0024] As shown in FIG. 2, the optimization device 10 includes an input / output storage 11 and an evaluation unit 12 .

[0025] As described above, the input / output storage 11 stores a plurality of input / output pairs {(I k,O k )} k=1~K is stored in advance.

[0026] The evaluation unit 12 evaluates a plurality of input / output pairs {(I k ,O k )} k=1~K In detail, the evaluation unit 12 optimizes the optimization target model 20 using a plurality of input / output pairs {(I k ,O k )} k=1~K The model 20 is optimized by using the above while evaluating the likelihood of the output signal y of the model 20 when an input signal x, which is an acoustic signal, is input to the model 20.

[0027] Next, the operation of the first embodiment will be described. FIG. 3 is a diagram illustrating a schematic operation example of the optimization device 10 according to the present disclosure. In FIG. 3, the input / output storage 11 stores a plurality of input / output pairs {(I k ,O k )} k=1~K It is also assumed that an output signal y of the model 20 when an input signal x, which is an acoustic signal, is input to the model 20 has already been obtained.

[0028] As shown in FIG. 3, first, the evaluation unit 12 evaluates a plurality of input / output pairs {(I k ,O k )} k=1~K is used to evaluate the likelihood of an output signal y of the model 20 when an input signal x, which is an acoustic signal, is input to the model 20 (step S11).

[0029] Next, the evaluation unit 12 optimizes the model 20 using the evaluation result of the likelihood of the output signal y of the model 20 (step S12). For example, if the likelihood of "1" is the best result, the evaluation unit 12 optimizes the model 20 by updating the parameters of the model 20 so that the likelihood of the output signal y approaches "1".

[0030] Here, the operation of step S11 in FIG. 3 will be described in detail. FIG. 4 is a flowchart illustrating an example of the operation flow of step S11 in FIG.

[0031] As shown in FIG. 4, first, the evaluation unit 12 evaluates a plurality of input / output pairs {(I k ,O k )} k=1~K Each input signal I k is compared with the output signal y of the model 20 (step S111). Next, the evaluation unit 12 evaluates a plurality of input / output pairs {(I k ,O k )} k=1~K Among them, the input signal I with the highest similarity to the output signal y of model 20 is selected. k Input / output pair (I k ,O k ) is extracted (step S112).

[0032] Next, the evaluation unit 12 evaluates the input / output pairs (I k ,O k ) output signal O k and the output signal y of the model 20 is calculated (step S113). Thereafter, the evaluation unit 12 outputs the calculation result of step S113 as an evaluation value of the likelihood of the output signal y of the model 20 (step S114).

[0033] That is, the evaluation unit 12 compares the output signal y of the model 20 with the input / output pair (I k ,O k ) output signal O k The closer the two are to each other, the more likely the output signal y of the model 20 is evaluated to be.

[0034] As described above, according to the first embodiment, the input / output storage 11 stores a plurality of input / output pairs {(I k ,O k )} k=1~K The evaluation unit 12 stores a plurality of input / output pairs {(I k ,O k)} k=1~K The optimization target model 20 is optimized using the above.

[0035] In this way, multiple input / output pairs {(I k ,O k )} k=1~K Therefore, when optimizing the target model 20, multiple input / output pairs {(I k ,O k )} k=1~K The required input / output pairs (I k ,O k ) to optimize the model 20. This makes it possible to reduce the amount of calculation required to optimize the model 20.

[0036] FIG. 5 is a diagram illustrating the effect of the optimization device 10 according to the present disclosure. In FIG. 5, the horizontal axis indicates optimization accuracy, which is the accuracy of the optimized model 20, with the optimization accuracy increasing as the value moves in the positive direction. The vertical axis indicates the amount of calculation required to optimize the model 20, with the amount of calculation increasing as the value moves in the positive direction. For comparison, FIG. 5 also shows the optimization accuracy and amount of calculation when the model 20 is optimized using a mathematical optimization method and when the model 20 is optimized using a related technology. Note that the optimization method for the model 20 using the related technology is, for example, a method of having the model 20 learn a machine learning model M, as described in Patent Document 1.

[0037] As shown in FIG. 5, the optimization method using the optimization device 10 can incorporate the statistical knowledge of the machine learning model M into the model 20, and therefore can achieve the same level of optimization accuracy as the optimization method using the related technology.

[0038] On the other hand, the optimization method using the optimization device 10 stores multiple input / output pairs of input signals and output signals of the machine learning model M in advance, and therefore the amount of calculation required for optimization can be reduced to the same amount of calculation as that required for mathematical optimization methods.

[0039] <Embodiment 2> First, the concept of the second embodiment will be described. FIG. 6 is a diagram illustrating the concept of an optimization device 10A according to the present disclosure.

[0040] As shown in FIG. 6, similar to the optimization device 10, the optimization device 10A is a device that incorporates statistical knowledge possessed by a machine learning model M into a model 20 to be optimized, and optimizes the model 20.

[0041] However, in this second embodiment, the machine learning model M is a trained model that uses an acoustic signal or a vibration signal observed by optical fiber sensing as an input signal and learns the noise components contained in the acoustic signal or the vibration signal. Also, the output signal of the machine learning model M is a signal in which noise has been suppressed from the input signal, which is an acoustic signal or a vibration signal.

[0042] For ease of explanation, the following description will be given assuming that the input signal of the machine learning model M is an acoustic signal that indicates the time series changes in the acoustics generated at points along the optical fiber, as observed by optical fiber sensing.

[0043] The optimization device 10A includes an input / output storage 11A. The input / output storage 11A, like the input / output storage 11, stores a plurality of input / output pairs {(I k ,O k )} k=1~K In addition, the input / output storage 11A stores in advance an input / output pair (second input / output pair) (I (n) ,O (n) ) is also stored in advance.

[0044] The optimization device 10A is configured to optimize a plurality of input / output pairs {(I k ,O k )} k=1~K and input / output pairs (I (n) ,O (n)) is used to optimize the model 20 while evaluating the likelihood, which indicates whether the output signal y of the model 20 is plausible (close to the desired solution) when an input signal x, which is an acoustic signal, is input to the model 20.

[0045] Next, the configuration of the second embodiment will be described. FIG. 7 is a block diagram showing a schematic configuration example of an optimization device 10A according to the present disclosure.

[0046] As shown in FIG. 7, an optimization device 10A includes an input / output storage 11A and an evaluation unit 12A.

[0047] As described above, the input / output storage 11A stores a plurality of input / output pairs {(I k ,O k )} k=1~K is stored in advance, and the input / output pair (I (n) ,O (n) ) is stored in advance.

[0048] The evaluation unit 12A evaluates a plurality of input / output pairs {(I k ,O k )} k=1~K and input / output pairs (I (n) ,O (n) ) to optimize the optimization target model 20. In detail, the evaluation unit 12A optimizes the optimization target model 20 by using a plurality of input / output pairs {(I k ,O k )} k=1~K and input / output pairs (I (n) ,O (n) ) is used to optimize the model 20 while evaluating the likelihood of the output signal y of the model 20 when an input signal x, which is an acoustic signal, is input to the model 20.

[0049] Next, the operation of the second embodiment will be described. FIG. 8 is a diagram illustrating a schematic operation example of the optimization device 10A according to the present disclosure. In FIG. 8, the input / output storage 11A stores a plurality of input / output pairs {(I k ,O k )} k=1~K and input / output pairs (I (n) ,O(n) ) has already been stored. It is also assumed that when an input signal x, which is an acoustic signal, is input to the model 20, an output signal y of the model 20 has already been obtained.

[0050] As shown in FIG. 8, first, the evaluation unit 12A evaluates a plurality of input / output pairs {(I k ,O k )} k=1~K and input / output pairs (I (n) ,O (n) ) is used to evaluate the likelihood of an output signal y of the model 20 when an input signal x, which is an acoustic signal, is input to the model 20 (step S21).

[0051] Next, the evaluation unit 12A optimizes the model 20 using the evaluation result of the likelihood of the output signal y of the model 20 (step S22). For example, if the likelihood of "1" is the best result, the evaluation unit 12A optimizes the model 20 by updating the parameters of the model 20 so that the likelihood of the output signal y approaches "1".

[0052] Here, the operation of step S21 in FIG. 8 will be described in detail. FIG. 9 is a flowchart illustrating an example of the operation flow of step S21 in FIG.

[0053] 9, first, the evaluation unit 12A performs the processes of steps S211 and S212, which are similar to steps S111 and S112 in FIG. 4. As a result, the evaluation unit 12A evaluates a plurality of input / output pairs {(I k ,O k )} k=1~K Among them, the input signal I with the highest similarity to the output signal y of model 20 is selected. k Input / output pair (I k ,O k ) to extract.

[0054] Next, the evaluation unit 12A evaluates the input / output pairs (I k ,O k ) output signal O k and the input / output pair (I (n) ,O(n) ) output signal O (n) The distance between and is calculated (step S213). Thereafter, the evaluation unit 12A outputs the calculation result of step S213 as an evaluation value of the likelihood of the output signal y of the model 20 (step S214).

[0055] That is, the evaluation unit 12A evaluates the input / output pairs (I k ,O k ) output signal O k and the input / output pair (I (n) ,O (n) ) output signal O (n) The closer the two are to each other, the more likely the output signal y of the model 20 is evaluated to be.

[0056] As described above, according to the second embodiment, the machine learning model M is a trained model that uses an acoustic signal or a vibration signal as an input signal to learn the noise components contained in the acoustic signal or the vibration signal, and it is assumed that the output signal of the machine learning model M is a signal in which noise has been suppressed from the acoustic signal or the vibration signal. Under this assumption, the input / output storage 11A stores a plurality of input / output pairs {(I k ,O k )} k=1~K and an input / output pair (I (n) ,O (n) The evaluation unit 12A stores a plurality of input / output pairs {(I k ,O k )} k=1~K and input / output pairs (I (n) ,O (n) ) to optimize the model 20 to be optimized.

[0057] In this way, multiple input / output pairs {(I k ,O k )}k=1~K and input / output pairs (I (n) ,O (n) ) are stored in advance, so when optimizing the target model 20, multiple input / output pairs {(I k ,O k )} k=1~K The required input / output pairs (I k ,O k ) and input / output pairs (I (n) ,O (n) ) and optimize the model 20. This makes it possible to reduce the amount of calculation required to optimize the model 20. In addition, since the statistical knowledge possessed by the machine learning model M can be incorporated into the model 20, it becomes possible to use the model 20 to suppress noise in the input signal, which is an acoustic signal.

[0058] <Third Embodiment> The third embodiment corresponds to an embodiment that is a higher-level conception of the first and second embodiments described above. FIG. 10 is a block diagram showing a schematic configuration example of an optimization device 10B according to the present disclosure.

[0059] As shown in FIG. 10, an optimization device 10B includes a storage unit 11B and an evaluation unit 12B.

[0060] The storage unit 11B pre-stores a plurality of first input / output pairs, which are pairs of input and output signals of the machine learning model M when an acoustic signal or vibration signal indicating a time series change in acoustics or vibration observed by optical fiber sensing is input to the machine learning model M as an input signal. The evaluation unit 12B optimizes the optimization target model 20 using a plurality of first input / output pairs.

[0061] In this way, since multiple input / output pairs of input signals and output signals of the machine learning model M are stored in advance, when optimizing the optimization target model 20, it is sufficient to use a necessary input / output pair from among the multiple input / output pairs to optimize the model 20. This makes it possible to reduce the amount of calculation required to optimize the model 20.

[0062] The evaluation unit 12B may compare the input signal of each of the multiple first input / output pairs with the output signal of the model 20 when an acoustic signal or a vibration signal is input as an input signal to the optimization target model 20, and extract from the multiple first input / output pairs a first input / output pair having an input signal that has the highest similarity to the output signal of the model 20. Furthermore, the evaluation unit 12B may evaluate the likelihood of the output signal of the model 20 using the output signal of the extracted first input / output pair and the output signal of the model 20, and optimize the model 20 using the evaluation result of the likelihood.

[0063] Furthermore, the machine learning model M may be a model that uses an acoustic signal or a vibration signal as an input signal to learn the relationship between the acoustic signal or the vibration signal and the text. Furthermore, the output signal of the machine learning model M may be text, or a combination of an acoustic signal or a vibration signal and text.

[0064] Furthermore, the machine learning model M may be a model that uses an acoustic signal or a vibration signal as an input signal and learns noise components contained in the acoustic signal or the vibration signal. Furthermore, the output signal of the machine learning model M may be a signal in which noise has been suppressed from an input signal that is an acoustic signal or a vibration signal.

[0065] In addition, the storage unit 11B may further pre-store a second input / output pair, which is a pair of an input signal and an output signal of the machine learning model M when an acoustic signal or a vibration signal with noise superimposed thereon is input to the machine learning model M as an input signal.

[0066] The evaluation unit 12B may also compare the input signal of each of the multiple first input / output pairs with the output signal of the model 20 when an acoustic signal or a vibration signal is input as an input signal to the optimization target model 20, and extract from the multiple first input / output pairs a first input / output pair having an input signal that has the highest similarity to the output signal of the model 20. Furthermore, the evaluation unit 12B may evaluate the likelihood of the output signal of the model 20 using the output signal of the extracted first input / output pair and the output signal of the second input / output pair, and optimize the model 20 using the evaluation result of the likelihood.

[0067] <Hardware configuration of the optimization device> FIG. 11 is a block diagram showing an example of a schematic hardware configuration of a computer 90 that realizes the optimization devices 10, 10A, and 10B.

[0068] 11, a computer 90 includes a processor 91, a memory 92, a storage 93, an input / output interface (input / output I / F) 94, and a communication interface (communication I / F) 95. The processor 91, the memory 92, the storage 93, the input / output interface 94, and the communication interface 95 are connected by a data transmission path for transmitting and receiving data to and from each other.

[0069] The processor 91 is, for example, an arithmetic processing device such as a CPU (Central Processing Unit) or a GPU (Graphics Processing Unit). The memory 92 is, for example, a memory such as a RAM (Random Access Memory) or a ROM (Read Only Memory). The storage 93 is, for example, a storage device such as an HDD (Hard Disk Drive), an SSD (Solid State Drive), or a memory card. The storage 93 may also be a memory such as a RAM or a ROM.

[0070] A program is stored in the storage 93. When the program is loaded into the computer, it includes a set of instructions (or software code) that causes the computer 90 to perform one or more functions of the optimization devices 10, 10A, and 10B described above. The components of the optimization devices 10, 10A, and 10B described above may be realized by the processor 91 reading and executing the program stored in the storage 93. Furthermore, the storage function of the optimization devices 10, 10A, and 10B described above may be realized by the memory 92 or the storage 93.

[0071] The above-described programs may also be stored on non-transitory computer-readable media or tangible storage media. By way of example and not limitation, computer-readable media or tangible storage media include RAM, ROM, flash memory, SSD or other memory technology, CD (Compact Disc)-ROM, DVD (Digital Versatile Disc), Blu-ray® disc or other optical disk storage, magnetic cassette, magnetic tape, magnetic disk storage or other magnetic storage devices. The programs may also be transmitted on a transitory computer-readable medium or communication medium. By way of example and not limitation, transitory computer-readable media or communication media include electrical, optical, acoustic, or other forms of propagated signals.

[0072] The input / output interface 94 is connected to a display device 941, an input device 942, a sound output device 943, etc. The display device 941 is a device that displays a screen corresponding to drawing data processed by the processor 91, such as an LCD (Liquid Crystal Display), a CRT (Cathode Ray Tube) display, or a monitor. The input device 942 is a device that accepts operational inputs from an operator, such as a keyboard, a mouse, or a touch sensor. The display device 941 and the input device 942 may be integrated and realized as a touch panel. The sound output device 943 is a device that outputs sound corresponding to audio data processed by the processor 91, such as a speaker.

[0073] The communication interface 95 transmits and receives data to and from an external device. For example, the communication interface 95 communicates with the external device via a wired communication path or a wireless communication path.

[0074] Although the present disclosure has been described above with reference to the embodiments, the present disclosure is not limited to the above-described embodiments. Various modifications that can be understood by those skilled in the art can be made to the configuration and details of the present disclosure within the scope of the present disclosure. Furthermore, each embodiment can be combined with other embodiments as appropriate.

[0075] Furthermore, each drawing is merely an example for describing one or more embodiments. Each drawing may relate not only to one particular embodiment, but also to one or more other embodiments. As will be understood by those skilled in the art, various features or steps described with reference to any one drawing can be combined with features or steps shown in one or more other drawings to create, for example, an embodiment not explicitly shown or described. Not all features or steps shown in any one drawing are necessarily required to describe an exemplary embodiment, and some features or steps may be omitted. The order of steps described in any drawing may be changed as appropriate.

[0076] Furthermore, some or all of the above-described embodiments can be described as, but are not limited to, the following supplementary notes. (Appendix 1) a storage means for storing in advance a plurality of first input / output pairs, which are pairs of input and output signals of a machine learning model when an acoustic signal or a vibration signal indicating a time-series change in acoustics or vibration observed by optical fiber sensing is input as an input signal to the machine learning model; and an evaluation means for optimizing a model to be optimized using the plurality of first input / output pairs. Optimizer. (Appendix 2) The evaluation means comparing the input signal of each of the plurality of first input / output pairs with an output signal of the optimization target model when the acoustic signal or the vibration signal is input as an input signal to the optimization target model; extracting a first input / output pair having an input signal that has the highest similarity to an output signal of the optimization target model from the plurality of first input / output pairs; using the extracted output signal of the first input / output pair and the output signal of the optimization target model, evaluate the likelihood of the output signal of the optimization target model; optimizing the optimization target model using the likelihood evaluation result; 2. The optimization apparatus of claim 1. (Appendix 3) The machine learning model is a model that uses the acoustic signal or the vibration signal as an input signal to learn a relationship between the acoustic signal or the vibration signal and text. 3. The optimization device according to claim 1 or 2. (Appendix 4) an output signal of the machine learning model is the text or a combination of the acoustic signal or the vibration signal and the text; 4. The optimization device of claim 3. (Appendix 5) The machine learning model is a model that uses the acoustic signal or the vibration signal as an input signal and learns noise components contained in the acoustic signal or the vibration signal. 2. The optimization apparatus of claim 1. (Appendix 6) The output signal of the machine learning model is a signal in which noise is suppressed from the input signal, which is the acoustic signal or the vibration signal. 6. The optimization device of claim 5. (Appendix 7) The storage means further stores in advance a second input / output pair, which is a pair of an input signal and an output signal of the machine learning model when the acoustic signal or the vibration signal on which noise is superimposed is input to the machine learning model as an input signal. 7. The optimization device of claim 6. (Appendix 8) The evaluation means comparing the input signal of each of the plurality of first input / output pairs with an output signal of the optimization target model when the acoustic signal or the vibration signal is input as an input signal to the optimization target model; extracting a first input / output pair having an input signal that has the highest similarity to an output signal of the optimization target model from the plurality of first input / output pairs; evaluating the likelihood of the output signal of the optimization target model using the extracted output signal of the first input-output pair and the output signal of the second input-output pair; optimizing the optimization target model using the likelihood evaluation result; 8. The optimization apparatus of claim 7. (Appendix 9) An optimization method executed by an optimization device, comprising: storing in advance a plurality of first input / output pairs, which are pairs of input and output signals of a machine learning model when an acoustic signal or a vibration signal indicating a time-series change in acoustics or vibration observed by optical fiber sensing is input as an input signal to the machine learning model; optimizing a model to be optimized using the plurality of first input / output pairs; Optimization methods. (Appendix 10) On the computer, a step of storing in advance a plurality of first input / output pairs, which are pairs of input and output signals of a machine learning model when an acoustic signal or a vibration signal indicating a time-series change in acoustics or vibration observed by optical fiber sensing is input as an input signal to the machine learning model; optimizing a model to be optimized using the plurality of first input / output pairs; A program that executes the following.

[0077] Note that some or all of the elements (e.g., configurations and functions) described in Supplementary Notes 2 to 8 that are dependent on Supplementary Note 1 may also be dependent on Supplementary Notes 9 and 10 in the same dependency relationship as Supplementary Notes 2 to 8. Some or all of the elements described in any Supplementary Note may be applied to various hardware, software, recording means for recording software, systems, and methods. [Explanation of symbols]

[0078] 10, 10A, 10B Optimization Device 11,11A Input / Output Storage 11B Storage Area 12, 12A, 12B Evaluation section 20 Model to be optimized 90 Computer 91 processors 92 memory 93 Storage 94 Input / Output Interface 941 Display device 942 Input Device 943 Sound Output Device 95 Communication Interface

Claims

1. a storage means for storing in advance a plurality of first input / output pairs, which are pairs of input and output signals of a machine learning model when an acoustic signal or a vibration signal indicating a time-series change in acoustics or vibration observed by optical fiber sensing is input as an input signal to the machine learning model; and an evaluation means for optimizing a model to be optimized using the plurality of first input / output pairs. Optimizer.

2. The evaluation means comparing the input signal of each of the plurality of first input / output pairs with an output signal of the optimization target model when the acoustic signal or the vibration signal is input as an input signal to the optimization target model; extracting a first input / output pair having an input signal that has the highest similarity to an output signal of the optimization target model from the plurality of first input / output pairs; evaluating a likelihood of the output signal of the optimization target model using the extracted output signal of the first input / output pair and the output signal of the optimization target model; optimizing the optimization target model using the likelihood evaluation result; The optimization device according to claim 1 .

3. The machine learning model is a model that uses the acoustic signal or the vibration signal as an input signal to learn a relationship between the acoustic signal or the vibration signal and text.

3. The optimization device according to claim 1 or 2.

4. The output signal of the machine learning model is the text or a combination of the acoustic signal or the vibration signal and the text. The optimization device according to claim 3 .

5. The machine learning model is a model that uses the acoustic signal or the vibration signal as an input signal and learns noise components contained in the acoustic signal or the vibration signal. The optimization device according to claim 1 .

6. The output signal of the machine learning model is a signal in which noise is suppressed from the input signal, which is the acoustic signal or the vibration signal. The optimization device according to claim 5 .

7. The storage means further stores in advance a second input / output pair, which is a pair of an input signal and an output signal of the machine learning model when the acoustic signal or the vibration signal on which noise is superimposed is input to the machine learning model as an input signal. The optimization device according to claim 6 .

8. The evaluation means comparing the input signal of each of the plurality of first input / output pairs with an output signal of the optimization target model when the acoustic signal or the vibration signal is input as an input signal to the optimization target model; extracting a first input / output pair having an input signal that has the highest similarity to an output signal of the optimization target model from the plurality of first input / output pairs; evaluating the likelihood of the output signal of the model to be optimized using the extracted output signal of the first input-output pair and the output signal of the second input-output pair; optimizing the optimization target model using the likelihood evaluation result; The optimization device according to claim 7 .

9. An optimization method executed by an optimization device, comprising: storing in advance a plurality of first input / output pairs, which are pairs of input and output signals of a machine learning model when an acoustic signal or a vibration signal indicating a time-series change in acoustics or vibration observed by optical fiber sensing is input as an input signal to the machine learning model; optimizing a model to be optimized using the plurality of first input / output pairs; Optimization methods.

10. On the computer, a step of storing in advance a plurality of first input / output pairs, which are pairs of input and output signals of a machine learning model when an acoustic signal or a vibration signal indicating a time-series change in acoustics or vibration observed by optical fiber sensing is input as an input signal to the machine learning model; optimizing a model to be optimized using the plurality of first input / output pairs; A program that executes the following.

Images