Abstract:Recent strides in low-latency spiking neural network (SNN) algorithms have drawn significant interest, particularly due to their event-driven computing nature and fast inference capability. One of the most efficient ways to construct a low-latency SNN is by converting a pre-trained, low-bit artificial neural network (ANN) into an SNN. However, this conversion process faces two main challenges: First, converting SNNs from low-bit ANNs can lead to ``occasional noise" -- the phenomenon where occasional spikes are generated in spiking neurons where they should not be -- during inference, which significantly lowers SNN accuracy. Second, although low-latency SNNs initially show fast improvements in accuracy with time steps, these accuracy growths soon plateau, resulting in their peak accuracy lagging behind both full-precision ANNs and traditional ``long-latency SNNs'' that prioritize precision over speed. In response to these two challenges, this paper introduces a novel technique named ``noise adaptor.'' Noise adaptor can model occasional noise during training and implicitly optimize SNN accuracy, particularly at high simulation times $T$. Our research utilizes the ResNet model for a comprehensive analysis of the impact of the noise adaptor on low-latency SNNs. The results demonstrate that our method outperforms the previously reported quant-ANN-to-SNN conversion technique. We achieved an accuracy of 95.95\% within 4 time steps on CIFAR-10 using ResNet-18, and an accuracy of 74.37\% within 64 time steps on ImageNet using ResNet-50. Remarkably, these results were obtained without resorting to any noise correction methods during SNN inference, such as negative spikes or two-stage SNN simulations. Our approach significantly boosts the peak accuracy of low-latency SNNs, bringing them on par with the accuracy of full-precision ANNs. Code will be open source.

Non-informative Noise-Enhanced Stochastic Neural Networks for Improving Adversarial Robustness

Enhance DNN Adversarial Robustness and Efficiency via Injecting Noise to Non-Essential Neurons

Exploiting Noise As a Resource for Computation and Learning in Spiking Neural Networks

A Comprehensive Analysis on Adversarial Robustness of Spiking Neural Networks

SNN-RAT: Robustness-enhanced Spiking Neural Network through Regularized Adversarial Training

May the Noise be with you: Adversarial Training without Adversarial Examples

Activity-dependent Neuron Model for Noise Resistance

Enhancing Adversarial Robustness in SNNs with Sparse Gradients

Improving Query Efficiency of Black-box Adversarial Attack

Parametric Noise Injection: Trainable Randomness to Improve Deep Neural Network Robustness against Adversarial Attack

Toward Robust Spiking Neural Network Against Adversarial Perturbation

Noise Adaptor in Spiking Neural Networks

Noise Optimization in Artificial Neural Networks

Training Robust Deep Neural Networks via Adversarial Noise Propagation

Artificial VO2 Spiking Neurons with Protective Mechanism for Enhancing Resilience of Spiking Neural Network Against Adversarial Attacks

Colored Noise Injection for Training Adversarially Robust Neural Networks

RSC-SNN: Exploring the Trade-off Between Adversarial Robustness and Accuracy in Spiking Neural Networks via Randomized Smoothing Coding

Adversarially robust neural networks with feature uncertainty learning and label embedding

Defense Against Query-Based Black-Box Attack with Small Gaussian-Noise

Noise Sensitivity-Based Energy Efficient and Robust Adversary Detection in Neural Networks

Neural SDE: Stabilizing Neural ODE Networks with Stochastic Noise