Neuron Module Learning Device for Spiking Neural Network Weight Updates

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing hardware, such as von Neumann-type devices, is less efficient in performing processes like pattern recognition, real-time recognition, and voice recognition compared to neuromorphic hardware, due to structural limitations like bandwidth.

Innovation Solution

A learning device for a neuron module in a spiking neural network (SNN) that includes a timer reset and restarted based on post-spikes, and a processor that determines post-then-pre and pre-then-post times to update synapse weights based on weight variations.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Productivity

If von Neumann-type hardware is used for pattern recognition and real-time recognition, then simple numerical calculations can be performed with desirable performance, but the processing efficiency is relatively low compared to neuromorphic hardware

Engineering Contradiction:
Improveprocessing efficiencyVSAvoidstructural limitations
Core Design Contradiction:
ProductivityVSDevice complexity

Solution Approach 1:

The patent replaces traditional von Neumann architecture with a neuromorphic system that mimics biological neural networks. The learning device uses spiking neural networks where neurons communicate through spikes, substituting the mechanical sequential processing of von Neumann hardware with a bio-inspired parallel event-driven architecture. This substitution enables more efficient pattern recognition and real-time processing by emulating the brain's natural information processing mechanisms.

Inventive Principle:
Principle #28Mechanics substitution (Replace mechanical system)

Solution Approach 2:

The patent implements dynamic parameter changes through spike-timing-dependent plasticity (STDP), where synapse weights are continuously adjusted based on the timing differences between pre-spike and post-spike events. This dynamic adaptation allows the system to learn and optimize its processing parameters in real-time, improving productivity for recognition tasks while managing structural complexity through adaptive rather than static configurations.

Inventive Principle:
Principle #35Parameter changes

2Productivity

If neuromorphic hardware is used to perform pattern recognition and real-time recognition, then processing efficiency is improved, but hardware complexity increases due to the need for specialized neural network structures

Engineering Contradiction:
Improverecognition processing efficiencyVSAvoidhardware structure
Core Design Contradiction:
ProductivityVSDevice complexity

Solution Approach 1:

The patent segments the neural network into discrete functional units: neurons that generate spikes, synapses that transmit and modulate signals, and a learning device that performs weight updates. This segmentation allows each component to be optimized independently for its specific function, improving overall recognition efficiency while managing hardware complexity through modular design. The timer is also segmented to track specific time intervals for STDP learning separately from general operation.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The learning device is designed as a universal platform that can implement various spiking neural network configurations and learning rules. The timer and processor work together to handle multiple types of temporal relationships (pre-then-post and post-then-pre times) and support different STDP curves, making the hardware adaptable to various recognition tasks without requiring task-specific redesign, thus improving productivity while controlling complexity through multi-functionality.

Inventive Principle:
Principle #6Universality (Multi-functionality)

3Adaptability or versatility

If synapse weights are dynamically updated based on timing information, then learning capability and recognition performance are improved, but computational complexity increases

Engineering Contradiction:
Improvelearning capabilityVSAvoidcomputational complexity
Core Design Contradiction:
Adaptability or versatilityVSDevice complexity

Solution Approach 1:

The patent implements periodic action through the timer that systematically tracks time intervals between spikes. The timer periodically resets and measures post-then-pre times and pre-then-post times according to spike events, creating a regular rhythm for weight update opportunities. This periodic structure simplifies the computational complexity by providing a predictable framework for when and how weights are updated, while still achieving strong learning capability through consistent temporal reinforcement.

Inventive Principle:
Principle #19Periodic action

Solution Approach 2:

The system implements self-service through autonomous STDP learning where the weight updates occur automatically based on spike timing without requiring external control. The timer and processor work together to automatically calculate time differences, determine appropriate weight variations based on STDP curves, and update synapse weights in real-time. This self-service mechanism improves adaptability and learning capability while reducing computational complexity by eliminating the need for complex external control systems.

Inventive Principle:
Principle #25Self-service

Data Source

PatentEP4567673A1Neuron module learning device and method of operating the same
Publication Date: 2025.06.11 SAMSUNG ELECTRONICS CO LTD
  • EP4567673A1 patent drawingFigure 1
  • EP4567673A1 patent drawingFigure 2
  • EP4567673A1 patent drawingFigure 3

AI summary

A learning device of a neuron module includes a timer configured to be reset and restarted, based on a post-spike occurring in the neuron module in a spiking neural network (SNN), and a processor configured to determine a post-then-pre time based on time information of the timer based on a pre-spike being received by at least one synapse of a plurality of synapses of the neuron module, determine a weight variation based on the post-then-pre time, and update a weight of the at least one synapse receiving the pre-spike, based on the weight variation.