Training Spiking Autoencoders by Truncated BPTT under Trade-offs between Simulation Steps and Reconstruction Error

Abstract

This article presents a comprehensive study of trade-offs between simulation steps and reconstruction performance for spiking autoencoders. We execute training and inference of a spiking neural network to reconstruct FashionMNSIT dataset images for several simulation step configurations and evaluate reconstruction accuracies by mean squared error. Experiments showed that a longer simulation step configuration indeed improves reconstruction accuracy while the improvement gets a peek at a very long configuration. Flexible design on the training configuration will be applicable; for example, shorter steps could be acceptable for accurate-insensitive but latency-restricted systems.