🤖 AI Summary
To address neural homeostasis disruption and low energy efficiency in spiking neural networks (SNNs) for event-based vision—caused by dynamic, asynchronous input streams—this work proposes the Asynchronous Biologically-plastic Neuron (ABN) mechanism. ABN introduces input-driven asynchronous dynamic thresholds and membrane potential decay, eliminating reliance on global clocks and fixed time steps, thereby enabling fully asynchronous, locally adaptive homeostasis maintenance. Built upon the leaky integrate-and-fire (LIF) model, it integrates asynchronous state updates, event-driven membrane evolution, and local synaptic plasticity. Evaluated across multiple event-camera datasets, ABN achieves 3.2–5.7% higher classification accuracy, a 4.1% improvement in semantic segmentation mIoU, a 38% reduction in neural activity entropy, and a 52% decrease in inference energy consumption. This work establishes a novel paradigm for robust, ultra-low-power neuromorphic vision processing.
📝 Abstract
Spiking Neural Networks (SNNs) offer a biologically inspired approach to computer vision that can lead to more efficient processing of visual data with reduced energy consumption. However, maintaining homeostasis within these networks is challenging, as it requires continuous adjustment of neural responses to preserve equilibrium and optimal processing efficiency amidst diverse and often unpredictable input signals. In response to these challenges, we propose the Asynchronous Bioplausible Neuron (ABN), a dynamic spike firing mechanism to auto-adjust the variations in the input signal. Comprehensive evaluation across various datasets demonstrates ABN's enhanced performance in image classification and segmentation, maintenance of neural equilibrium, and energy efficiency.
