Spiking neural networks improve image classification with multi-depth temporal fusion
Multi-Depth Temporal Fusion for Feedforward, Locally Trained Spiking Neural Networks
Neural and Evolutionary ComputingComputer Vision and Pattern RecognitionMachine Learning
Summary
The paper explores how spiking neural networks (SNNs), which mimic brain-like processing using spike timing, can better recognize images by combining information from different network layers over time. The authors designed a new method called multi-depth temporal fusion that keeps early signals and adds later ones only if they agree in timing. This approach helps the network learn locally and efficiently, achieving strong results on popular image tasks like MNIST and CIFAR-10 without needing complex training methods. It also uses fewer spikes, making the network more energy efficient.
What this means in practice
- •For embedded systems engineers: Design energy-efficient image recognition modules for devices by using locally trained spiking neural networks with multi-depth temporal fusion to reduce spike processing.
- •For neuromorphic hardware developers: Implement multi-layer spiking neural networks with temporal fusion mechanisms to improve real-time visual data classification using sparse spike events.$Commercial implications: This enables development of advanced neuromorphic processors that perform image recognition more accurately and with lower power consumption.
Authors
Aidin Attar, Eleonora Cicciarella, Michele Rossi
Abstract
We propose a new spiking neural network (SNN) design to process static images and event streams using time-to-first-spike (TTFS) latencies. Our key research question is which architectural choices best accommodate local and online learning in multi-layer convolutional SNNs. This question is addressed via an original framework combining residual-like connections with multi-depth feature aggregation and consensus. The full SNN pipeline features an early-vision front end, to convert raw visual data into sparse spike latencies, a four-layer convolutional backbone trained layerwise with unsupervised spike-timing-dependent plasticity (STDP), a deterministic Multi-Depth Temporal Fusion (MDTF) and a final classifier trained with reward-modulated spike-timing-dependent plasticity (R-STDP). Rather than replacing early features in deeper layers, the proposed MDTF preserves early temporal evidence, adding sparse residual events from intermediate layers, and incorporating deeper features only when they agree in time with earlier representations. The resulting architecture is experimentally validated across MNIST, Fashion-MNIST, CIFAR-10, and N-MNIST, delivering strong classification performance under a fully local learning regime. Selective multi-depth fusion significantly outperforms traditional STDP/R-STDP baselines on higher-variability visual tasks (achieving +18.2 pp on Fashion-MNIST and +29.2 pp on CIFAR-10). Furthermore, activity-budget analyses show that the network retains high accuracy even when removing a large fraction of late or weak spike events, confirming its high data efficiency and reduced event-processing requirements. The codebase is publicly available at github.com/aidinattar/multi-depth- temporal-fusion-snn.