US-12619860-B2 - Method and device for controlling firing timing in spiking neural networks

US12619860B2US 12619860 B2US12619860 B2US 12619860B2US-12619860-B2

Abstract

A computation apparatus that includes a spiking neuron model. A spiking neuron model varies an index value of a signal output based on an input condition of a signal during an input time interval and outputs, based on the index value, a signal during an output time interval that starts after the input time interval ends.

Inventors

Yusuke SAKEMI
Takeo Hosomi

Assignees

NEC CORPORATION

Dates

Publication Date: 20260505
Application Date: 20220725
Priority Date: 20210806

Claims (7)

1 . A computation apparatus that comprises: one or more memories storing instructions; and one or more processors configured to execute the instructions to control the computation apparatus to implement a spiking neuron model that: varies a potential of a signal output based on an input condition of a signal during an input time interval; outputs, based on the potential, a signal during an output time interval that starts after the input time interval ends; varies the potential at a variation speed in accordance with the input condition of the signal during the input time interval; when the potential does not attain a threshold value during the input time interval, varies the potential at a variation speed that is same as the variation speed at end of the input time interval; when the potential attains the threshold value during the input time interval, outputs the signal at start of the output time interval; when the potential attains the threshold value during the output time interval, outputs the signal when the potential attains the threshold value; and when the potential does not attain the threshold value during the output time interval, outputs the signal at end of the output time interval.
2 . The computation apparatus according to claim 1 , wherein the spiking neuron model comprises a first neuron model that outputs the signal and a second neuron model that receives the signal from the first neuron model, and the output time interval and the input time interval are set so that the output time interval of the first neuron model overlaps with the input time interval of the second neuron model.
3 . The computation apparatus according to claim 1 , further comprising a clock circuit configured to generate a clock signal, wherein the spiking neuron model detects the input time interval and the output time interval based on the clock signal.
4 . A computation method comprising: varying an index value of a signal output based on an input condition of a signal during an input time interval; outputting, based on the index value, a signal during an output time interval that starts the input time interval ends; varying the index value at a variation speed in accordance with the input condition of the signal during the input time interval; when the index value does not attain a threshold value during the input time interval, varying the index value at a variation speed that is same as the variation speed at end of the input time interval; when the index value attains the threshold value during the input time interval, outputting the signal at start of the output time interval; when the index value attains the threshold value during the output time interval, outputting the signal when the index value attains the threshold value; and when the index value does not attain the threshold value during the output time interval, outputting the signal at end of the output time interval.
5 . The computation method according to claim 4 , further comprising: controlling a clock circuit to generate a clock signal; and detecting the input time interval and the output time interval based on the clock signal.
6 . A non-transitory recording medium that stores a program that causes a programmable apparatus to execute: varying an index value of a signal output based on an input condition of a signal during an input time interval; outputting, based on the index value, a signal during an output time interval that starts after the input time interval ends; varying the index value at a variation speed in accordance with the input condition of the signal during the input time interval; when the index value does not attain a threshold value during the input time interval, varying the index value at a variation speed that is same as the variation speed at end of the input time interval; when the index value attains the threshold value during the input time interval, outputting the signal at start of the output time interval; when the index value attains the threshold value during the output time interval, outputting the signal when the index value attains the threshold value; and when the index value does not attain the threshold value during the output time interval, outputting the signal at end of the output time interval.
7 . The non-transitory recording medium according to claim 6 , wherein the program further causes the programmable apparatus to execute: controlling a clock circuit to generate a clock signal; and detecting the input time interval and the output time interval based on the clock signal.

Description

This application is based upon and claims the benefit of priority from Japanese patent application No. 2021-130003, filed on Aug. 6, 2021, the disclosure of which is incorporated herein in its entirety by reference. TECHNICAL FIELD The present disclosure relates to a computation apparatus, a neural network system, a neuron model apparatus, a computation method and a program. BACKGROUND ART One type of neural network is a spiking neural network (SNN). For example, Japanese Unexamined Patent Application, First Publication No. 2018-136919 A (hereinafter Patent Document 1) describes a neuromorphic computing system in which a spiking neural network is implemented on a neuromorphic computing device. In the spiking neural network, neuron models have internal states known as membrane potentials, and they output signals known as spikes based on the time evolution of the membrane potentials. SUMMARY A spiking neural network should preferably be able to efficiently perform data processing. An example objective of the present disclosure is to provide a computation apparatus, a neural network system, a neuron model apparatus, a computation method and a program that can solve the above-mentioned problems. According to a first example aspect of the present disclosure, a computation apparatus includes: a spiking neuron model that: varies an index value of a signal output based on an input condition of a signal during an input time interval; and outputs, based on the index value, a signal during an output time interval that starts after the input time interval ends. According to a second example aspect of the present disclosure, a spiking neural network system includes a spiking neural network body and learning means. The neural network body includes neuron models including index value calculation means for varying an index value of a signal output based on an input condition of a signal during an input time interval; and signal output means for outputting, based on the index value, a signal during an output time interval that starts after the input time ends. The learning means learns a weighting coefficient for the signal. According to a third example aspect of the present disclosure, a spiking neuron model apparatus includes index value calculation means for varying an index value of a signal output based on an input condition of a signal during an input time interval; and signal output means for outputting, based on the index value, a signal during an output time interval that starts after the input time interval ends. According to a fourth example aspect of the present disclosure, a computation method includes: varying an index value of a signal output based on an input condition of a signal during an input time interval; and outputting, based on the index value, a signal during an output time interval that starts after the input time interval ends. According to a fifth example aspect of the present disclosure, a non-transitory recording medium stores a program that causes a programmable apparatus to execute: varying an index value of a signal output based on an input condition of a signal during an input time interval; and outputting, based on the index value, a signal during an output time interval that starts after the input time interval ends. According to the present disclosure, a spiking neural network can efficiently perform data processing. BRIEF DESCRIPTION OF THE DRAWINGS FIG. 1 is a diagram illustrating an example of the structure of a neural network apparatus according to an example embodiment. FIG. 2 is a diagram illustrating an example of the structure of a spiking neural network provided in a neural network apparatus according to the example embodiment. FIG. 3 is a diagram illustrating an example of the time variation in the membrane potential in a spiking neuron model in which the output timings of spike signals are not restricted according to the example embodiment. FIG. 4 is a diagram illustrating an example of the output timings of spike signals from spiking neuron models in a spiking neural network in the case in which the output timings of the spike signals from the spiking neuron models are not restricted according to the example embodiment. FIG. 5 is a diagram illustrating an example of spike signal transfer timings between neuron models in a neural network apparatus according to the example embodiment. FIG. 6 is a diagram illustrating an example of the setting of time intervals according to the example embodiment. FIG. 7 is a diagram illustrating a first example of a clip function according to the example embodiment. FIG. 8 is a diagram illustrating a second example of a clip function according to the example embodiment. FIG. 9 is a diagram illustrating an example of the system structure when learning according to the example embodiment. FIG. 10 is a diagram illustrating an example of input and output of signals in a neural network system according to the example embodiment. FIG. 11 is a diagram