- This event has passed.
Yifan Gong PhD Dissertation Defense
August 9, 2024 @ 11:00 am - 12:00 pm
Name:
Yifan Gong
Title:
Towards Energy-Efficient Deep Learning for Sustainable AI
Date:
8/9/2024
Time:
11:00:00 AM
Committee Members:
Prof. Yanzhi Wang (Advisor)
Prof. David R. Kaeli
Prof. Xue Lin
Prof. Huaizu Jiang
Prof. Stratis Ioannidis
Abstract:
The rapid advancements in deep learning (DL) and artificial intelligence (AI) have led to transformative applications across various domains, such as community virtual reality experiences, autonomous systems, and climate change prediction. Edge devices including mobile and embedded systems play a vital role in carrying these applications, facilitating the widespread adoption of machine intelligence. Along with the great success of DL and AI is the huge energy consumption for both training and inference. With the breakthrough of large-scale models for AI-generated content (AIGC) such as large language models and diffusion models, the energy consumption issue intensifies, causing the urgent need for sustainable AI solutions. In this talk, I will talk about how to facilitate deep learning on various edge devices in an energy-efficient manner for the goal of sustainable AI. Specifically, I will start by introducing my two system-level approaches to tackling the challenge. The first approach is named bottom-up, which conducts AI algorithm-aware efficient system design. The second approach is a top-down approach that achieves hardware-driven efficient AI algorithm design. Then, I will share my recent works addressing the efficiency issues for large-scale models. Finally, I will show the applications of my methods and pointers to the future direction.