# US 2021/0232910 A1 · Reservoir computing

> Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt

- Number: US 2021/0232910 A1
- Document ID: US20210232910A1
- Status: Published application
- Kind: A1
- Application number: 16/775,008
- Filed: 2020-01-28
- Published: 2021-07-29
- Earliest claimed: 2020-01-28
- Assignees: International Business Machines Corporation
- Applicants: International Business Machines Corporation
- Inventors: Seiji Takeda; Toshiyuki Yamane; Daiju Nakano
- Web page: https://patentgrep.com/patent/US20210232910A1

## CPC

- G06N 3/08: Learning methods
- G01N 21/65: Raman scattering
- G06F 18/2148: {characterised by the process organisation or structure, e.g. boosting cascade}
- G06F 18/2185: {the supervisor being an automated module, e.g. intelligent oracle}
- G06F 18/24133: {Distances to prototypes}
- G06N 3/044: Recurrent networks, e.g. Hopfield networks
- G06N 3/0675: {using electro-optical, acousto-optical or opto-electronic means}
- G06N 3/09: Supervised learning
- G06V 10/7747: {Organisation of the process, e.g. bagging or boosting}
- G06V 10/7788: {the supervisor being a human, e.g. interactive learning with a human teacher}
- G06V 10/82: using neural networks
- H01S 5/06804: {by monitoring an external parameter, e.g. temperature}
- H03M 1/12: Analogue/digital converters ({H03M1/001 – } H03M1/10 take precedence)
- G06N 3/048: Activation functions
- G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G01N: INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES (measuring or testing apparatus or processes other than immunoassay, involving enzymes or microorganisms C12M, C12Q)
- G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N)
- G06V: IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- H01S: DEVICES USING THE PROCESS OF LIGHT AMPLIFICATION BY STIMULATED EMISSION OF RADIATION [LASER] TO AMPLIFY OR GENERATE LIGHT; DEVICES USING STIMULATED EMISSION OF ELECTROMAGNETIC RADIATION IN WAVE RANGES OTHER THAN OPTICAL
- H03M: CODING; DECODING; CODE CONVERSION IN GENERAL (using fluidic means F15C4/00; optical analogue/digital converters G02F7/00; coding, decoding or code conversion, specially adapted for particular applications, see the relevant subclasses, e.g. G01D, G01R, G06F, G06T, G09G, G10L, G11B, G11C, H04B, H04L, H04M, H04N; ciphering or deciphering for cryptography or other purposes involving the need for secrecy G09C)

## Abstract

Provided is a reservoir computing system including a reservoir having a random laser for emitting a non-linear optical signal with respect to an input signal. The reservoir computing system also includes a converter for converting the non-linear optical signal into an output signal by applying a conversion function. The conversion function is trained by using a training input signal and a target output signal.

## Claims

### Claim 1

A reservoir computing system comprising:
a reservoir including a random laser for emitting a non-linear optical signal with respect to an input signal; and
a converter for converting the non-linear optical signal into an output signal by applying a conversion function.

### Claim 2 (depends on 1)

The reservoir computing system of claim 1, wherein the conversion function is trained by using a training input signal and a target output signal.

### Claim 3 (depends on 1)

The reservoir computing system of claim 1, wherein the input signal is an electrical signal.

### Claim 4 (depends on 1)

The reservoir computing system of claim 1, wherein the input signal is an optical signal.

### Claim 5 (depends on 1)

The reservoir computing system of claim 1, wherein the output signal is an electrical signal.

### Claim 6 (depends on 5)

The reservoir computing system of claim 5, wherein the converter includes:
a photo detector for converting the non-linear optical signal into an electrical signal;
an AD converter for converting the electrical signal detected by the photo detector into a digital signal; and
a digital signal converter for converting the digital signal from the AD converter into the output signal by applying the conversion function to the digital signal.

### Claim 7 (depends on 6)

The reservoir computing system of claim 6, wherein the converter includes a plurality of photo detectors receiving non-linear optical signals each of which are emitted from a different area of the random laser.

### Claim 8 (depends on 6)

The reservoir computing system of claim 6, wherein the digital signal converter calculates a weighted sum of time-series values of the digital signal using a plurality of trained weights.

### Claim 9 (depends on 1)

The reservoir computing system of claim 1, further comprising a training apparatus for training the conversion function of the converter by using the training input signal and the target output signal.

### Claim 10 (depends on 9)

The reservoir computing system of claim 9, further comprising a bias unit for adding bias to the input signal to be input to the random laser, and
wherein the training apparatus trains the bias by using the training input signal and the target output signal.

### Claim 11 (depends on 1)

The reservoir computing system of claim 1, further comprising a plurality of instances of the random laser, and
wherein the converter includes a reservoir selector for selecting a random laser, from among the plurality of instances of the random laser, that is to be used to emit the non-linear optical signal to be converted to the output signal.

### Claim 12 (depends on 11)

The reservoir computing system of claim 11, wherein the training apparatus trains the reservoir selector to select a random laser, from among the plurality of instances of the random laser, based on the accuracy of each random laser.

### Claim 13 (depends on 1)

The reservoir computing system of claim 1, further comprising a temperature control unit for controlling a temperature of the random laser.

### Claim 14

A method comprising:
supplying an input signal corresponding to an object to a reservoir including a random laser for emitting a non-linear optical signal with respect to the input signal;
converting the non-linear optical signal into an output signal by applying a conversion function; and
identifying the object by using the output signal.

### Claim 15

A training apparatus comprising:
an obtaining unit for obtaining a training input signal and a target output signal;
a supplying unit for supplying the training input signal to a reservoir computing system, wherein the reservoir computing system includes a random laser for emitting a non-linear optical signal with respect to an input signal and a converter for converting the non-linear optical signal into an output signal by applying a conversion function; and
a training unit for training the conversion function to reduce an error between the output signal and the target output signal corresponding to the training input signal.

### Claim 16 (depends on 15)

The training apparatus of claim 15, wherein
the reservoir computing system further includes a bias unit for adding bias to the input signal to be input to the random laser, and
the training unit further trains the bias by using the training input signal and the target output signal.

### Claim 17 (depends on 15)

The training apparatus of claim 15, wherein
the reservoir computing system further includes a plurality of instances of the random laser,
the converter includes a reservoir selector for selecting a random laser, from among the plurality of instances of the random laser, that is to be used to emit the non-linear optical signal to be converted to the output signal, and
the training unit trains the reservoir selector to select a random laser from among the plurality of instances of the random laser based on the accuracy of each random laser.

### Claim 18 (depends on 15)

The training apparatus of claim 15, wherein the reservoir receives the input signal corresponding to an object and the converter outputs the output signal to represent an identification of the object.

### Claim 19

A computer-implemented method comprising:
obtaining a training input signal and a target output signal;
supplying the training input signal to a reservoir computing system, wherein the reservoir computing system includes a random laser for emitting a non-linear optical signal with respect to an input signal and a converter for converting the non-linear optical signal to an output signal by applying a conversion function; and
training the conversion function to reduce an error between the output signal and the target output signal corresponding to the training input signal.

### Claim 20 (depends on 19)

The computer-implemented method of claim 19, wherein
the reservoir computing system further includes a bias unit for adding bias to the input signal to be input to the random laser, and
the training includes training the bias by using the training input signal and the target output signal.

### Claim 21 (depends on 19)

The computer-implemented method of claim 19, wherein
the reservoir computing system further includes a plurality of instances of the random laser,
the converter includes a selector for selecting a random laser, from among the plurality of instances of the random laser, that is to be used to emit the non-linear optical signal to be converted to the output signal, and
the training includes training the reservoir selector to select a random laser from among the plurality of instances of the random laser based on the accuracy of each random laser.

### Claim 22 (depends on 19)

The computer-implemented method of claim 19, wherein the reservoir receives the input signal corresponding to an object and the converter outputs the output signal to represent an identification of the object.

### Claim 23

A computer program product including one or more computer readable storage mediums collectively storing program instructions that are executable by a processor or programmable circuitry to cause the processor or programmable circuitry to perform operations comprising:
obtaining a training input signal and a target output signal;
supplying the training input signal to a reservoir computing system, wherein the reservoir computing system includes a random laser for emitting non-linear optical signal with respect to an input signal and a converter for converting the non-linear optical signal to an output signal by applying a conversion function; and
training the conversion function to reduce an error between the output signal and the target output signal corresponding to the training input signal.

### Claim 24 (depends on 23)

The computer program product of claim 23, wherein
the reservoir computing system further includes a bias unit for adding bias to the input signal to be input to the random laser, and
the training includes further training the bias by using the training input signal and the target output signal.

### Claim 25 (depends on 23)

The computer program product of claim 23, wherein the reservoir computing system further includes a plurality of instances of the random laser and the converter includes a reservoir selector for selecting a random laser, from among the plurality of instances of the random laser, that is to be used to emit the non-linear optical signal to be converted to the output signal, and
the training includes further training the reservoir selector to select a random laser from among the plurality of instances of the random laser based on the accuracy of each random laser.
