StartUps general


    AI-DA Challenge


    CSS Master Programme


    Energy Optimization

Faculty News

Accurate and efficient time-domain classification with adaptive spiking recurrent neural networks

September 14th, 2021

Inspired by more detailed modeling of biological neurons, Spiking neural networks (SNNs) have been investigated both as more biologically plausible and potentially more powerful models of neural computation, and also with the aim of extracting biological neurons’ energy efficiency; the performance of such networks however has remained lacking compared to classical artificial neural networks (ANNs). Here, we demonstrate how a novel surrogate gradient combined with recurrent networks of tunable and adaptive spiking neurons yields state-of-the-art for SNNs on challenging benchmarks in the time-domain, like speech and gesture recognition. This also exceeds the performance of standard classical recurrent neural networks (RNNs) and approaches that of the best modern ANNs. As these SNNs exhibit sparse spiking, we show that theoretically they are one to three orders of magnitude more computationally efficient compared to RNNs with comparable performance.
Together, we argue that this positions SNNs as an attractive solution for AI hardware implementations.

Prof. Dr. Sander M. Bohté

Seminarraum IGI - Inffeldgasse 16b/I

Friday, 17 September 2021 11.00 – 12.45


Contact us

Dean's Office
Inffeldgasse 10/II, 8010 Graz

Office Hours:
Mon 10:00 - 11:00 and 14:00 - 16:00
Tue to Fri 09:00 - 12:00

Online consultation on Discord!


Durch laden des Tweeds werden ihre Daten an Twitter übermittlet. Es gelten die Datenschutzbestimmungen von Twitter.