BiSNN: Training Spiking Neural Networks with Binary Weights via Bayesian Learning
This work addresses the problem of energy-efficient inference for battery-powered devices by combining binary weights and sparse binary activations in SNNs, offering an incremental improvement for embedded AI.
This paper introduces a Spiking Neural Network (SNN) model that utilizes both temporally sparse binary activations and binary weights to enhance energy efficiency. The authors derive two learning rules, with the Bayesian paradigm demonstrating an advantage in accuracy and calibration compared to full-precision implementations.
Artificial Neural Network (ANN)-based inference on battery-powered devices can be made more energy-efficient by restricting the synaptic weights to be binary, hence eliminating the need to perform multiplications. An alternative, emerging, approach relies on the use of Spiking Neural Networks (SNNs), biologically inspired, dynamic, event-driven models that enhance energy efficiency via the use of binary, sparse, activations. In this paper, an SNN model is introduced that combines the benefits of temporally sparse binary activations and of binary weights. Two learning rules are derived, the first based on the combination of straight-through and surrogate gradient techniques, and the second based on a Bayesian paradigm. Experiments validate the performance loss with respect to full-precision implementations, and demonstrate the advantage of the Bayesian paradigm in terms of accuracy and calibration.