Neuromorphic Architecture With Internal State Neuron Information

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Conventional neuromorphic computing systems face limitations in learning multiple correlations and reliability, particularly due to their reliance on inhibitory links which hinder effective data feature insight and performance.

Innovation Solution

A neuromorphic architecture featuring interconnected electronic neurons with internal state information links, enabling the modification of neuron operations and enhancing data processing capabilities through spike-timing dependent plasticity synapses and spiking neural networks.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Ease of operation

If inhibitory links are used in neuromorphic circuits, then the circuit can implement winner-take-all configuration, but the performance in learning multiple correlations deteriorates

Engineering Contradiction:

Improvewinner-take-all configurationVSAvoidlearning multiple correlations

Core Design Contradiction:

Ease of operationVSAdaptability or versatility

Solution Approach 1:

The patent inverts the conventional inhibitory link approach by using excitatory links instead. Rather than neurons inhibiting each other to achieve winner-take-all, the system uses excitatory connections where neurons mutually enhance each other's activity, allowing multiple neurons to remain active simultaneously and learn multiple correlations effectively

Inventive Principle:

Principle #13The other way round (Inversion)

Solution Approach 2:

The patent implements feedback mechanisms where the internal state information of neurons is fed back to modify synaptic weights dynamically. This feedback loop allows the system to adapt and learn multiple correlations by continuously adjusting connection strengths based on ongoing neural activity patterns

Inventive Principle:

Principle #23Feedback

2Ease of manufacture

If conventional CMOS transistor technology and Von Neumann architectures are used, then computing elements can be implemented, but power requirements increase significantly

Engineering Contradiction:

Improvecomputing element implementationVSAvoidpower requirements

Core Design Contradiction:

Ease of manufactureVSUse of energy by moving object

Solution Approach 1:

The patent replaces conventional CMOS transistor-based computing elements with neuromorphic circuit elements that mimic biological neuron behavior. These neuromorphic elements use event-driven spiking mechanisms rather than continuous voltage levels, significantly reducing power consumption while maintaining computational functionality

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

Solution Approach 2:

The patent employs periodic spiking action potentials as the fundamental computing mechanism. Instead of continuous operation, neurons fire discrete spikes at specific intervals when thresholds are reached, creating an event-driven system that consumes power only during active computation events rather than continuously

Inventive Principle:

Principle #19Periodic action

3Measurement precision

If internal state information links are used to interconnect neurons, then the ability to detect correlations and recognize patterns improves, but the device complexity increases

Engineering Contradiction:

Improvecorrelation detection capabilityVSAvoidneuron interconnection structure

Core Design Contradiction:

Measurement precisionVSDevice complexity

Solution Approach 1:

The patent makes the internal state information links serve multiple functions simultaneously: they transmit neuron state information, provide feedback for synaptic weight modification, enable correlation detection, and support pattern recognition. This multi-functionality reduces the need for separate dedicated circuits for each function, thereby managing complexity while enhancing capability

Inventive Principle:

Principle #6Universality (Multi-functionality)

Data Source

PatentUS10755167B2Neuromorphic architecture with multiple coupled neurons using internal state neuron information

Publication Date: 2020.08.25 INTERNATIONAL BUSINESS MACHINE CORPORATION

AI summary

This invention relates to an apparatus, system, and method for computing with neuromorphic circuit architectures that have neurons with interconnected internal state information. The interconnected internal state information allows the neurons to enable or strengthen the input to other neurons. Furthermore, neuron internal state information provides insights on the characteristics of the input data that can be used to enhance the performance of the neuromorphic system. The neuromorphic system can be implemented with an artificial phase-change-based neurons.