Title: Energy Efficient and Secure Design of Deep Learning for the Internet of Things
Dr. Mukhopadhyay, Advisor
Dr. Krishna, Chair
The objective of the proposed research is to design an energy efficient and secure deep learning system for the Internet of Things (IoTs). The research particularly focuses on energy efficient training of deep learning for online learning or domain adaptation on edge devices, and robust secure deep learning on the clouds. To enable energy efficient training of deep learning, the research studies impact of a limited precision training of various types of neural networks like convolutional neural networks (CNNs) and recurrent neural networks (RNNs). For CNNs, the work proposes dynamic precision scaling algorithm, and precision flexible computing unit to accelerate CNNs training. For RNNs, the work studies impact of various hyper-parameters to enable low precision training of RNNs and proposes low precision computing unit with stochastic rounding. For secure deep learning, the research proposes adversarial machine learning regularized with a unified embedding for image classification and low level (pixel level) similarity learning. Development of an online training algorithm for domain adaptation on the edge devices will complete the research, enabling practical system design of an energy-efficient and secure deep learning for the IoTs.