Spiking Neural Network Image Recognition via STDP Learning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image recognition devices based on neural networks face challenges with high computational complexity and heat generation due to error backpropagation learning, making them difficult to apply to spiking neural networks.
Innovation Solution
An image recognition device using a brain-inspired spiking neural network with a spiking neural network unit that includes neurons corresponding to image pixels, synapses connecting them, and employs Spike Timing-Dependent plasticity (STDP) learning to modify synaptic weights, along with firing rate and spike timing coding for neural encoding, to recognize images.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If error backpropagation learning is used in neural networks, then learning capability is improved, but computational complexity and heat generation increase
Solution Approach 1:
The patent replaces the mechanical/mathematical error backpropagation system with a biologically-inspired STDP learning mechanism. Instead of using gradient descent and weight updates based on error gradients, the system uses spike-timing-dependent plasticity where synaptic weights are modified based on the relative timing of pre-synaptic and post-synaptic spikes, fundamentally substituting the learning mechanism to reduce computational complexity
Solution Approach 2:
The patent changes the learning parameters from continuous weight values updated via error gradients to discrete spike timing events. By parameterizing learning around spike times rather than continuous error signals, the system achieves learning capability with reduced computational burden and lower heat generation
2Reliability
If error backpropagation learning is used in neural networks, then learning capability is improved, but heat generation increases
Solution Approach 1:
The patent substitutes the high-energy error backpropagation mechanism with a low-energy STDP mechanism that operates on spike timing events. This replacement eliminates the need for complex matrix operations and gradient calculations that generate significant heat, while maintaining learning capability through biologically-plausible synaptic plasticity
3Adaptability or versatility
If spiking neural networks are used, then biological plausibility is improved, but ease of implementation deteriorates due to difficulty in applying learning rules
Solution Approach 1:
The patent implements self-service learning through STDP where the network automatically adjusts synaptic weights based on intrinsic spike timing patterns without requiring external error signals or complex training algorithms. The system serves itself by using its own activity patterns to drive learning, making spiking neural networks both biologically plausible and easier to implement
Data Source
AI summary
Disclosed are an image recognition device and method using a brain-inspired spiking neural network and a computer-readable program for the same. The image recognition device using a brain-inspired spiking neural network according to the present disclosure includes an input unit configured to receive an input image made up of at least one pixel, and a spiking neural network unit configured to recognize the input image, the spiking neural network unit including a plurality of neurons corresponding to the pixels of the image to generate spike signals when a membrane potential state value exceeds a preset threshold, and synapses connecting the plurality of neurons.