Hardware Lookup Table Synapse Weight Update for Neuromorphic STDP
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional software-based Spike Time Dependent Plasticity (STDP) neuromorphic systems face challenges in real-time on-system learning due to time-consuming calculations and limited resolution in updating synapse weight values, which hinder efficient implementation of on-chip self-learning functions.
Innovation Solution
A hardware-based method using synapse weight incrementers and decrementers with lookup tables to generate updated synapse weight values responsive to spike timing data, integrated into a neuromorphic system for improved simulation performance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If software-based calculation is used to update synapse weight values, then the system can implement STDP model with simple mathematical model, but the computation time is too long to achieve real-time on-chip learning
Solution Approach 1:
The patent replaces software-based mathematical calculations with hardware-based circuits (incrementers, decrementers, lookup tables) to perform synapse weight updates. This substitution of computational mechanism achieves real-time processing speeds while maintaining the STDP model functionality, resolving the contradiction between model simplicity and computation speed.
2Measurement precision
If conventional software program is used to calculate synapse weight values, then the system can update weights, but the update resolution cannot be increased much due to simple digital pulse shape
Solution Approach 1:
The patent changes the representation parameters of synapse weights from simple digital pulses to multi-bit resolution values stored in lookup tables. This parameter change enables higher update resolution (e.g., 8-bit or 16-bit precision) while the hardware circuits handle the complexity, separating precision enhancement from program complexity.
3Ease of operation
If conventional scheme is used to calculate and load synapse weight values, then the system can update weights, but real-time on-chip learning cannot be implemented due to long calculation and loading time
Solution Approach 1:
The patent pre-calculates and stores synapse weight update values in lookup tables during system initialization or offline preparation. During runtime, the hardware circuits simply retrieve and apply these pre-computed values, eliminating real-time calculation delays and enabling on-chip self-learning without time loss.
4Productivity
If hardware-based approach is used to update synapse weights, then real-time on-chip learning can be achieved, but the system requires special hardware circuits for SW update function
Solution Approach 1:
The patent segments the synapse weight update function into separate hardware modules: incrementer circuits for weight increase, decrementer circuits for weight decrease, and lookup tables for storing update values. This segmentation allows each component to be simple and specialized, reducing overall hardware complexity while achieving real-time processing capability.
Data Source
AI summary
A method and system are provided for updating synapse weight values in neuromorphic system with Spike Time Dependent Plasticity model. The method includes selectively performing, by a hardware-based synapse weight incrementer or decrementer, one of a synapse weight increment function or decrement function, each using a respective lookup table, to generate updated synapse weight values responsive to spike timing data. The method further includes storing the updated synapse weight values in a memory. The method additionally includes performing, by a hardware-based processor, a learning process to integrate the updated synapse weight values stored in the memory into the Spike Time Dependent Plasticity model neuromorphic system for improved neuromorphic simulation.