A Spiking Neural Network with Visible Light Communications

摘要

Spiking neural networks benefit from the real-time response when implemented using the analogue hardware since each electronic neuron operates independently from each other. The main problem of such neural networks is when the neural areas are non-static and are at a significant distance from each other. One solution to solve this problem is to use wireless based connectivity with parallel information transmission capability between neurons. This work presents the design and implementation of an all-optical synapse, based on the wavelength division multiplexed visible light communications between neurons, which has been adopted to overcome the problem of real-time weights adjustment. The results showed that by direct conversion of the electronic spikes into optical pulses, the spiking neurons are able to communicate through the optical channel with very low influence on the neural network normal spiking activity.