Temporal coding in leaky spiking neural networks

Abstract

Spiking neural networks that perform temporal encoding for phase-coherent neural computing are provided. In particular, according to an aspect of the present disclosure, a spiking neural network can include one or more spiking neurons that have an activation layer that uses a double exponential function to model a leaky input that an incoming neuron spike provides to a membrane potential of the spiking neuron. The use of the double exponential function in the neuron's temporal transfer function creates a better defined maximum in time. This allows very clearly defined state transitions between"now" and the "future step" to happen without loss of phase coherence.

Description

[0001] related application [0002] This application claims priority to and benefit from U.S. Provisional Patent Application No. 62 / 744,150, filed October 11, 2018. US Provisional Patent Application No. 62 / 744,150 is hereby incorporated by reference in its entirety. technical field [0003] The present disclosure generally relates to neural networks. More specifically, the present disclosure relates to leaky spiking neural networks that perform temporal encoding. Background technique [0004] Traditionally, artificial neural networks have been constructed primarily from idealized neurons that generate continuous activation values ​​based on a set of weighted inputs, using non-linear activation layers. Some neural networks have multiple sequential layers of such neurons, in which case these neural networks may be referred to as "deep" neural networks. [0005] Non-spiking neural networks typically use non-linear activation layers that produce continuous-valued outputs to t...

AI Technical Summary

Problems solved by technology

Therefore, these methods cannot be used to train multi-layer networks

[0011] Training especially spiking networks with multilayer learning (e.g., deep spiking neural networks) remains a challenge

Images