Cross-Point Device using Ta2O5/Ta Layer for Synapse Element in Neural Network

Abstract

We have developed a cross-point device using a Ta2O5/Ia layer for synapse elements in neural networks. Horizontal 80 and vertical 80 platinum lines and a Ta2O5/Ia layer between them make 6400 cross-point synapses integrated on a glass substrate. The electrical conductance gradually degrades by flowing current, which is available for modified Hebbian learning. We found that a neural network using the cross-point synapses are able to recognize multiple letters.