Juan Sanchez-Vazquez

Major and Classification

Electrical Engineering

Faculty Mentor

Rehan Kapadia


Viterbi School of Engineering

Research Gateway Project

“Machine Vision: Machine Learning at the Hardware Level”

Project Abstract

aults in the von Neumann computer architecture as well as approaching physical limitations in CMOS computing have renewed interest in alternative computer architectures. That in addition to the ubiquitous use of machine learning and artificial neural networks in pattern recognition can benefit from neuromorphic computing. As significant developments have been made in emulating brain-like activity in neuromorphic devices, it is important to observe potential system-level learning for the future of neuromorphic devices. A photodiode array image sensor artificial neural network is demonstrated. By applying a cross-entropy backpropagation algorithm, the image sensor achieved classification accuracy above 90% after 8 training epochs. This demonstration serves as a proof-of-concept supporting the possibility of using an array structure artificial neural network to simultaneously store and process data.