Abstract:Spiking Neural Networks (SNNs) are considered as a potential candidate for the next generation of artificial intelligence with appealing characteristics such as sparse computation and inherent temporal dynamics. By adopting architectures of Artificial Neural Networks (ANNs), SNNs achieve competitive performances on benchmark tasks like image classification. However, successful architectures of ANNs are not optimal for SNNs. In this work, we apply Neural Architecture Search (NAS) to find suitable architectures for SNNs. Previous NAS methods for SNNs focus primarily on the spatial dimension, with a notable lack of consideration for the temporal dynamics that are of critical importance for SNNs. Drawing inspiration from the heterogeneity of biological neural networks, we propose a differentiable approach to optimize SNN on both spatial and temporal dimensions. At spatial level, we have developed a spike-based differentiable hierarchical search (SpikeDHS) framework, where spike-based operation is optimized on both the cell and the layer level under computational constraints. We further propose a differentiable surrogate gradient search (DGS) method to evolve local SG functions independently during training. At temporal level, we explore an optimal configuration of diverse temporal dynamics on different types of spiking neurons by evolving their time constants, based on which we further develop hybrid networks combining SNN and ANN, balancing both accuracy and efficiency. Our methods achieve comparable classification performance of CIFAR10/100 and ImageNet with accuracies of 96.43%, 78.96%, and 70.21%, respectively. On event-based deep stereo, our methods find optimal layer variation and surpass the accuracy of specially designed ANNs with 26$\times$ lower computational cost ($6.7\mathrm{mJ}$), demonstrating the potential of SNN in processing highly sparse and dynamic signals.

Learning Spatial and Temporal Information with Complex-Valued Operation for High-Performance of Spiking Neural Networks

RSNN: Recurrent Spiking Neural Networks for Dynamic Spatial-Temporal Information Processing

A Brain-Inspired Spiking Neural Network Model with Temporal Encoding and Learning

Deep Spiking Neural Network Using Spatio-temporal Backpropagation with Variable Resistance.

Hierarchical Spiking-Based Model for Efficient Image Classification with Enhanced Feature Extraction and Encoding.

Deep CovDenseSNN: A Hierarchical Event-Driven Dynamic Framework with Spiking Neurons in Noisy Environment

Toward Efficient Processing and Learning with Spikes: New Approaches for Multispike Learning

CSNN: an Augmented Spiking Based Framework with Perceptron-Inception

Pattern Recognition Computation in a Spiking Neural Network with Temporal Encoding and Learning

Rapid Feedforward Computation by Temporal Encoding and Learning with Spiking Neurons

Efficient Spiking Neural Networks with Sparse Selective Activation for Continual Learning

Adaptive Multi-Level Firing for Direct Training Deep Spiking Neural Networks

Exploiting Neuron and Synapse Filter Dynamics in Spatial Temporal Learning of Deep Spiking Neural Network

IM-LIF: Improved Neuronal Dynamics with Attention Mechanism for Direct Training Deep Spiking Neural Network

Ultra-low Latency Spiking Neural Networks with Spatio-Temporal Compression and Synaptic Convolutional Block

Fast and Accurate Classification with a Multi-Spike Learning Algorithm for Spiking Neurons.

Spiking neural network with working memory can integrate and rectify spatiotemporal features

Spatiotemporal Backpropagation based on Channel Reward for Training High-Precision Spiking Neural Network

Spatial-Temporal Search for Spiking Neural Networks

Spatio-Temporal Backpropagation for Training High-Performance Spiking Neural Networks

Supervised Learning in Multilayer Spiking Neural Networks With Spike Temporal Error Backpropagation