Shrestha, S. B., Zhu, L., & Sun, P. (2022). Spikemax: Spike-based Loss Methods for Classification. 2022 International Joint Conference on Neural Networks (IJCNN). https://doi.org/10.1109/ijcnn55064.2022.9892379
Abstract:
Spiking Neural Networks (SNNs) are a promising
research paradigm for low power edge-based com-
puting. Recent works in SNN backpropagation has
enabled training of SNNs for practical tasks. How-
ever, since spikes are binary events in time, stan-
dard loss formulations are not directly compati-
ble with spike output. As a result, current works
are limited to using mean-squared loss of spike
count. In this paper, we formulate the output prob-
ability interpretation from the spike count measure
and introduce spike-based negative log-likelihood
measure which are more suited for classification
tasks. We compare our loss measures with other
existing alternatives and evaluate using classifica-
tion performances on three neuromorphic bench-
mark datasets: NMNIST, DVS Gesture and N-
TIDIGITS18. In addition, we demonstrate state of
the art performances on these datasets.
License type:
Publisher Copyright
Funding Info:
This research / project is supported by the National Research Foundation - Research, Innovation and Enterprise 2020 plan (Advanced Manufacturing and Engineering domain)
Grant Reference no. : A1687b0033