HIGHLIGHTS
- What: The authors propose a method to reduce inference energy and latency in SNNs based on Bayesian fusion. The authors use the integrate-and-fire (IF) model, which is considered to be the most popular model and has been proposed in many hardware implementations. The authors provide the method to transform the early layer activations into the prior knowledge. To alleviate this problem, the authors propose to incorporate the prior knowledge of the spike firing probability with the observed spike train to improve the estimation accuracy.
- Who: Takehiro Habara from the Centre National de la . . .
If you want to have access to all the content you need to log in!
Thanks :)
If you don't have an account, you can create one here.