We gratefully acknowledge support from
the Simons Foundation and member institutions.
Full-text links:


Current browse context:


Change to browse by:

References & Citations

DBLP - CS Bibliography


(what is this?)
CiteULike logo BibSonomy logo Mendeley logo del.icio.us logo Digg logo Reddit logo ScienceWISE logo

Computer Science > Neural and Evolutionary Computing

Title: Sparse Spiking Gradient Descent

Abstract: There is an increasing interest in emulating Spiking Neural Networks (SNNs) on neuromorphic computing devices due to their low energy consumption. Recent advances have allowed training SNNs to a point where they start to compete with traditional Artificial Neural Networks (ANNs) in terms of accuracy, while at the same time being energy efficient when run on neuromorphic hardware. However, the process of training SNNs is still based on dense tensor operations originally developed for ANNs which do not leverage the spatiotemporally sparse nature of SNNs. We present here the first sparse SNN backpropagation algorithm which achieves the same or better accuracy as current state of the art methods while being significantly faster and more memory efficient. We show the effectiveness of our method on real datasets of varying complexity (Fashion-MNIST, Neuromophic-MNIST and Spiking Heidelberg Digits) achieving a speedup in the backward pass of up to 150x, and 85% more memory efficient, without losing accuracy.
Subjects: Neural and Evolutionary Computing (cs.NE); Emerging Technologies (cs.ET); Machine Learning (cs.LG); Neurons and Cognition (q-bio.NC)
Journal reference: Advances in Neural Information Processing Systems (NeurIPS, 2021)
Cite as: arXiv:2105.08810 [cs.NE]
  (or arXiv:2105.08810v2 [cs.NE] for this version)

Submission history

From: Nicolas Perez-Nieves [view email]
[v1] Tue, 18 May 2021 20:00:55 GMT (15649kb,D)
[v2] Thu, 13 Jan 2022 17:07:56 GMT (18536kb,D)

Link back to: arXiv, form interface, contact.