Neurosynaptic Structural Description Framework for Neuronal Activity Simulation

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Current neuromorphic and synaptronic computation systems lack effective methods to simulate and control neuronal activity in artificial neural networks, particularly in routing neuronal firing events and implementing spike-timing dependent plasticity (STDP) for long-term potentiation and depression.

Innovation Solution

The development of a structural description framework for neurosynaptic core circuits with interconnect networks of electronic synapses, which simulates desired neuronal activity by programming core circuits and controls the routing of neuronal firing events through a composer and decomposer framework, utilizing input and output mapping tables to manage inputs and outputs across multiple interconnected core circuits.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If traditional digital models are used for neural network computation, then routing and control of neuronal firing events becomes straightforward, but the system fails to effectively simulate biological neuronal activity and implement STDP

Engineering Contradiction:

Improvesimulation accuracy of neuronal activityVSAvoidrouting and control of neuronal firing events

Core Design Contradiction:

ReliabilityVSEase of operation

Solution Approach 1:

The patent introduces a structural description framework as an intermediary layer between the hardware implementation and the simulation requirements. This framework includes composer and decomposer tools that translate high-level neural network descriptions into detailed routing configurations, enabling both accurate STDP simulation and ease of operation without direct manual routing control

Inventive Principle:

Principle #24Intermediary (Mediator)

2Manufacturing precision

If manual routing control is implemented for neuronal firing events, then routing precision can be improved, but the complexity of programming and controlling neural networks increases significantly

Engineering Contradiction:

Improverouting precision of neuronal firing eventsVSAvoidcomplexity of programming neural networks

Core Design Contradiction:

Manufacturing precisionVSDevice complexity

Solution Approach 1:

The patent implements preliminary action by providing structural description templates and composition frameworks that predefine routing patterns and STDP implementations. The composer tool automatically generates detailed routing configurations from high-level specifications, eliminating the need for manual routing programming while maintaining precision

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The framework uses copying by allowing neural network architectures to be described through reusable structural templates. The structural description can be copied and adapted across different implementations, with the decomposer automatically instantiating specific routing configurations from these templates, reducing programming complexity

Inventive Principle:

Principle #26Copying

3Measurement precision

If detailed structural descriptions are created for each core circuit, then simulation accuracy of desired neuronal activity improves, but the time and effort required to program the neural network increases

Engineering Contradiction:

Improveprecision of desired neuronal activity simulationVSAvoidtime to program neural network

Core Design Contradiction:

Measurement precisionVSLoss of time

Solution Approach 1:

The patent segments the neural network programming task into hierarchical levels: high-level structural descriptions, intermediate composition specifications, and low-level implementation details. The composer automatically handles the translation between levels, allowing users to work at higher abstraction levels while maintaining precise simulation control through the segmented framework

Inventive Principle:

Principle #1Segmentation

4Reliability

If the system implements accurate STDP and synaptic conductance control, then biological fidelity of the neural network improves, but the complexity of the interconnect network and routing fabric increases

Engineering Contradiction:

Improvebiological fidelity of neural networkVSAvoidcomplexity of interconnect network

Core Design Contradiction:

ReliabilityVSDevice complexity

Solution Approach 1:

The patent implements universality by designing a unified structural description framework that handles multiple functions: defining neuronal activity patterns, specifying STDP rules, configuring synaptic conductance, and generating routing configurations. This single framework manages the complexity of biological fidelity requirements without proportionally increasing interconnect network complexity

Inventive Principle:

Principle #6Universality (Multi-functionality)

Data Source

PatentUS11151444B2Structural descriptions for neurosynaptic networks

Publication Date: 2021.10.19 INTERNATIONAL BUSINESS MACHINE CORPORATION

AI summary

Embodiments of the invention provide a method comprising creating a structural description for at least one neurosynaptic core circuit. Each core circuit comprises an interconnect network including plural electronic synapses for interconnecting one or more electronic neurons with one or more electronic axons. The structural description defines a desired neuronal activity for the core circuits. The desired neuronal activity is simulated by programming the core circuits with the structural description. The structural description controls routing of neuronal firing events for the core circuits.