Title: Theoretical Error Performance Analysis for Deep Neural Networks based Regression Functional Approximation
Committee:
Dr. Lee, Advisor
Dr. Ma, Co-Advisor
Dr. Anderson, Chair
Dr. Romberg
Dr. Juang
Abstract: The objective of the proposed research is to exploit statistical learning and nonconvex optimization to analyze deep neural network (DNN) based vector-tovector regression by providing DNN with upper bounds on representation and generalization powers.