SkipSNN: Efficiently Classifying Spike Trains with Event-attention

TL;DR

This paper introduces an event-attention mechanism that enables SNNs to dynamically highlight useful signals of the original spike trains and proposes SkipSNN, which extends existing SNN models by learning to mask out noise by skipping membrane potential updates and shortening the effective size of the computational graph.

Abstract

Spike train classification has recently become an important topic in the machine learning community, where each spike train is a binary event sequence with \emph{temporal-sparsity of signals of interest} and \emph{temporal-noise} properties. A promising model for it should follow the design principle of performing intensive computation only when signals of interest appear. So such tasks use mainly Spiking Neural Networks (SNNs) due to their consideration of temporal-sparsity of spike trains. However, the basic mechanism of SNNs ignore the temporal-noise issue, which makes them computationally expensive and thus high power consumption for analyzing spike trains on resource-constrained platforms. As an event-driven model, an SNN neuron makes a reaction given any input signals, making it difficult to quickly find signals of interest. In this paper, we introduce an event-attention mechanism that enables SNNs to dynamically highlight useful signals of the original spike trains. To this end, we propose SkipSNN, which extends existing SNN models by learning to mask out noise by skipping membrane potential updates and shortening the effective size of the computational graph. This process is analogous to how people choose to open and close their eyes to filter the information they see. We evaluate SkipSNN on various neuromorphic tasks and demonstrate that it achieves significantly better computational efficiency and classification accuracy than other state-of-the-art SNNs.

Figures (6)

• Figure 1: The problem definition of efficient classification of spike trains. The spike trains are generated by an event camera, which is an imaging sensor that responds to local changes in brightness. Each pixel inside an event camera operates independently and asynchronously, reporting changes in brightness as they occur, and staying silent otherwise. Therefore, each image can be considered as binary event image.
• Figure 2: Differences among Recurrent Neural Network (RNN) , SkipRNN campos2017skip, Spiking Neural Network (SNN) wu2018spatioshrestha2018slayer and SkipSNN (ours). SkipSNN outperforms others in computational efficiency and classification accuracy with an event-attention mechanism for noise filtering.
• Figure 3: Model architecture of the proposed SkipSNN.
• Figure 4: Observing the performance of different models with different percentages of updated time-steps.
• Figure 5: Temporal raster plot of spikes in the controller neuron during the inference of the examples in N-MNIST and DVS-Gesture by SkipSNN. The spikes shown are generated from the controller, and used to control when to enter awake state. The blue box represents the period of useful signals. The dash red line represents the location of awake state.