Bridging the Gap between ANNs and SNNs by Calibrating Offset Spikes
This work addresses a key bottleneck for deploying SNNs on neuromorphic chips by improving accuracy at low latency, though it is incremental as it builds on existing ANN-SNN conversion methods.
The paper tackles the performance degradation of Spiking Neural Networks (SNNs) under low time-steps in ANN-SNN conversion by defining an offset spike to measure deviation and proposing an optimization strategy based on shifting initial membrane potential, achieving state-of-the-art performance with a top-1 accuracy of 67.12% on ImageNet using 6 time-steps.
Spiking Neural Networks (SNNs) have attracted great attention due to their distinctive characteristics of low power consumption and temporal information processing. ANN-SNN conversion, as the most commonly used training method for applying SNNs, can ensure that converted SNNs achieve comparable performance to ANNs on large-scale datasets. However, the performance degrades severely under low quantities of time-steps, which hampers the practical applications of SNNs to neuromorphic chips. In this paper, instead of evaluating different conversion errors and then eliminating these errors, we define an offset spike to measure the degree of deviation between actual and desired SNN firing rates. We perform a detailed analysis of offset spike and note that the firing of one additional (or one less) spike is the main cause of conversion errors. Based on this, we propose an optimization strategy based on shifting the initial membrane potential and we theoretically prove the corresponding optimal shifting distance for calibrating the spike. In addition, we also note that our method has a unique iterative property that enables further reduction of conversion errors. The experimental results show that our proposed method achieves state-of-the-art performance on CIFAR-10, CIFAR-100, and ImageNet datasets. For example, we reach a top-1 accuracy of 67.12% on ImageNet when using 6 time-steps. To the best of our knowledge, this is the first time an ANN-SNN conversion has been shown to simultaneously achieve high accuracy and ultralow latency on complex datasets. Code is available at https://github.com/hzc1208/ANN2SNN_COS.