- Keywords: RNNs, Spiking neurons, Neuromorphics
- TL;DR: A technique to train spiking RNNs to achieve high accuracy without simulating spikes.
- Abstract: The increasing need for compact and low-power computing solutions for machine learning applications has triggered a renaissance in the study of energy-efficient neural network accelerators. In particular, in-memory computing neuromorphic architectures have started to receive substantial attention from both academia and industry. However, most of these architectures rely on spiking neural networks, which typically perform poorly compared to their non-spiking counterparts in terms of accuracy. In this paper, we propose a new adaptive spiking neuron model that can also be abstracted as a low-pass filter. This abstraction enables faster and better training of spiking networks using back-propagation, without simulating spikes. We show that this model dramatically improves the inference performance of a recurrent neural network and validate it with three complex spatio-temporal learning tasks: the temporal addition task, the temporal copying task, and a spoken-phrase recognition task. Application of these results will lead to the development of powerful spiking models for neuromorphic hardware that solve relevant edge-computing and Internet-of-Things applications with high accuracy and ultra-low power consumption.