Zambrano & Bohte; - Fast and Efficient Asynchronous Neural Computation with Adapting Spiking)

Motivation

We know that real neurons are very efficient units; they encode input signal and communicate information to downstream neurons with only few spikes. However, it remains unclear how they manage to do that exactly.

The question is can we learn from biology to create similarly highly efficient networks doing computation?

Sparsity of spikes in a network is a particularly appealing feature for always-on applications, ones running on low-power budget.

To explain how neurons encode a signal with few spikes ;efficiently;. A recent line of works suggest that neurons implement a form of online AD/DA conversion;as as Yoon’s paper.