Programmable Neurons for Analog Nonvolatile Memory in Deep Learning Artificial Neural Networks

摘要

A number of embodiments are disclosed for processing the current output of a vector-by-matrix multiplication (VMM) array in an artificial neural network. Embodiments include a summer circuit and an activation function circuit. The summer circuit and/or activation function circuit includes circuit elements that can be adjusted in response to the total possible current received from the VMM to optimize power consumption.