Neuron Model Time Interval Segmentation for Spiking Neural Network Throughput

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Spiking neural networks face inefficiencies in data processing due to unrestricted output timing of spike signals, leading to delays in processing multiple sets of data sequentially.

Innovation Solution

Implementing a computing device with a neuron model that divides input and output time intervals for spike reception and transmission, where the neuron model fires within the output time interval and restricts firing during the input time interval, allowing for synchronized processing across layers.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Speed

If the neuron model allows unrestricted output timing for spike signals, then the neuron model can respond immediately to input signals, but this causes delays in processing multiple sets of data sequentially and reduces processing efficiency

Engineering Contradiction:

Improveresponse speedVSAvoiddata processing throughput

Core Design Contradiction:

SpeedVSProductivity

Solution Approach 1:

The patent divides the time interval into input time interval and output time interval. The input time interval is the period when the neuron model receives input signals, and the output time interval is the period when the neuron model outputs spike signals. By segmenting time this way, the system can receive multiple input signals during the input time interval without generating output signals until the output time interval, enabling efficient sequential processing of multiple data sets.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The patent implements preliminary action by accumulating input signals during the input time interval before generating output signals in the output time interval. The neuron model integrates input signals over the input time interval and only generates spike outputs after this accumulation period, allowing the system to prepare and process multiple data sets efficiently without immediate output delays.

Inventive Principle:

Principle #10Preliminary action

2Productivity

If the neuron model restricts firing during the input time interval, then data processing efficiency and throughput are improved, but this requires precise time interval management and control mechanisms

Engineering Contradiction:

Improvedata processing throughputVSAvoidtime interval control mechanism

Core Design Contradiction:

ProductivityVSDevice complexity

Solution Approach 1:

The patent applies dynamics by making the neuron model's firing behavior time-dependent. The neuron model dynamically changes its state based on the current time interval: during the input time interval, it receives signals without firing; during the output time interval, it generates spike signals based on accumulated input. This dynamic state change is controlled by timing mechanisms that switch between input and output modes, achieving efficient processing without requiring complex continuous control.

Inventive Principle:

Principle #15Dynamics

Data Source

PatentUS20240370712A1Computing device, neural network system, neuron model device, computing method, and program

Publication Date: 2024.11.07 NEC CORP

AI summary

A computing device includes: a neuron model in which an input time interval for receiving a spike and an output time interval for firing and a transmitting spike are divided, wherein the neuron model fires within the output time interval, with firing within the input time interval being restricted.