ECE PhD Dissertation Defense: Ning Liu

PhD Dissertation Defense: Real-World Applicable Deep Learning Techniques: From Efficient Modeling to Automated Model Optimization

Ning Liu

Location: Zoom Link

Abstract: Recently, deep neural networks (DNNs) have been widely studied and achieved tremendous success in a variety of real-world applications, such as computer vision, medical diagnosis and machine translation. Deep reinforcement learning (DRL), as an emerging powerful deep learning technique, combines DNNs with reinforcement learning into an interactive system. DRL opens up many new applications in domains such as healthcare, robotics and smart grids. With the rapid evolution of IT infrastructures, cloud computing has been witnessed as the prevailing computing paradigm. The underlying infrastructure of cloud computing relies on a large amount of data centers. The energy efficiency issue from “cloud” becomes more crucial and calls for more attentions.
In this dissertation, to solve the real-world energy efficiency problems, we take advantage of the deep learning and deep reinforcement learning techniques for efficient modeling of “cloud” applications. We present a DNN-based power management framework for regulation service and a novel DRL-based hierarchical framework for solving the overall resource allocation and power management problem. On the other hand, the powerful DNNs themselves are massive, consuming tremendous energy. Therefore, we explore the efficiency on deep neural networks. We propose an automatic model pruning framework to reduce the storage and computation requirements and accelerate inference. Our framework outperforms the prior work on automatic model compression by up to 33× in pruning rate (120× reduction in the actual parameter count) under the same accuracy. Significant inference speedup has been observed from the proposed framework on actual measurements on smartphone.