RMP-SNN: Residual Membrane Potential Neuron for Enabling Deeper High-Accuracy and Low-Latency Spiking Neural Network

By Bing Han, Gopalakrishnan Srinivasan, Kaushik Roy. Computer Vision and Pattern Recognition, 2020

411 citations recorded by the source corpus.

Spiking Neural Networks (SNNs) have recently attracted significant research interest as the third generation of artificial neural networks that can enable low-power event-driven data analytics. The best performing SNNs for image recognition tasks are obtained by converting a trained Analog Neural Network (ANN), consisting of Rectified Linear Units (ReLU), to SNN composed of integrate-and-fire neurons with "proper" firing thresholds. The converted SNNs typically incur loss in accuracy compared to that provided by the original ANN and require sizable number of inference time-steps to achieve the best accuracy. We find that performance degradation in the converted SNN stems from using "hard reset" spiking neuron that is driven to fixed reset potential once its membrane potential exceeds the firing threshold, leading to information loss during SNN inference. We propose ANN-SNN conversion using "soft reset" spiking neuron model, referred to as Residual Membrane Potential (RMP) spiking neuron, which retains the "residual" membrane potential above threshold at the firing instants. We demonstrate near loss-less ANN-SNN conversion using RMP neurons for VGG-16, ResNet-20, and ResNet-34 SNNs on challenging datasets including CIFAR-10 (93.63% top-1), CIFAR-100 (70.93% top-1), and ImageNet (73.09% top-1 accuracy). Our results also show that RMP-SNN surpasses the best inference accuracy provided by the converted SNN with "hard reset" spiking neurons using 2-8 times fewer inference time-steps across network architectures and datasets.

Subfields: Cellular and Molecular Neuroscience, Cognitive Neuroscience, Electrical and Electronic Engineering

View Paper: https://www.semanticscholar.org/paper/8aad18f5c474cd25e283a7b3c956894d94f82873

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Cellular and Molecular Neuroscience

In Cognitive Neuroscience

In Electrical and Electronic Engineering

Citations in the corpus

Spikformer: When Spiking Neural Network Meets Transformer (2022, methodology): “The converted SNN requires large time steps to accurately approximate ReLU activation, which causes large latency (Han et al., 2020).”
Going Deeper With Directly-Trained Larger Spiking Neural Networks (2020, background): “The ﬁrst one is called as the “ANN-SNN conversion meth-ods” (Sengupta et al. 2019; Han, Srinivasan, and Roy 2020), seen as the most popular way to train deep SNNs with high performance, which transforms the real-valued output of ReLU function to binary spikes in the SNN.”
DIET-SNN: A Low-Latency Spiking Neural Network With Direct Input Encoding and Leakage and Threshold Optimization (2021, methodology): “Most of the recent work on supervised learning in SNNs has either employed the IF or the LIF neuron model [13], [14], [18], [23], [24].”
Deep Residual Learning in Spiking Neural Networks (2021, methodology): “ANN to SNN conversion (ANN2SNN) [20, 4, 46, 49, 12, 11, 6, 54, 33] and backpropagation with surrogate gradient [40] are the two main methods to get deep SNNs.”
Temporal Efficient Training of Spiking Neural Network via Gradient Re-weighting (2022, methodology): “Some special techniques have been proposed to reduce the inference latency, such as the subtraction mechanism (Rueckauer et al., 2016; Han et al., 2020), robust normalization Rueckauer et al. (2016), spike-norm (Sengupta et al., 2018), and channel-wise normalization (Kim et al., 2019).”
Incorporating Learnable Membrane Time Constant to Enhance Learning of Spiking Neural Networks (2020, methodology): “Dataset SOTA SOTA’s T ours T MNIST [64] 400 8 Fashion-MNIST [8] 20 8 CIFAR-10 [19] 2048 8 N-MNIST [58] 59-64 10 CIFAR10-DVS [58] 230-292 20”