 |
Speaker: Peizhong Ju
Date: Feb 21, 11:45am–12:45pm Abstract: Machine Learning (ML), a vital branch of Artificial Intelligence (AI), has seen rapid advancements in recent years. As ML continues to evolve, it faces The first part of my presentation delves into the theoretical understanding of overparameterization, a hallmark of modern ML models like Deep Neural Networks (DNNs) and Large Language Models (LLMs). I will begin by exploring the performance of basis pursuit (min L1-norm) in linear regression models within the overparameterized region. This will lead to a discussion on our findings related to 2-layer and 3-layer Neural Tangent Kernel (NTK) models. Subsequently, I will present insights into multi-task learning scenarios, such as meta-learning and continual learning, under the lens of overparameterization. The second part of the presentation shifts focus to the application of ML in edge devices. Here, I will highlight the unique challenges and opportunities that arise in resource-constrained environments. We will explore strategies to tackle issues like computational complexity and fairness, which are crucial for the effective deployment of AI technologies in edge computing scenarios. Biographical Sketch: Peizhong Ju has been a postdoctoral researcher at the ECE department of Ohio State University since 2021, supervised by Ness Shroff. He received his B.S. degree in Electrical Engineering from Peking University in China in 2016. He obtained his Ph.D. degree in Electrical and Computer Engineering at Purdue University in 2021, advised by Xiaojun Lin. His research interests include machine learning, smart grid, and wireless communication. His research has received Best Paper Award in ACM e-Energy 2022, spotlight presentation in NeurIPS 2020, and Best Paper Award Finalist in ACM e-Energy 2018. Location and Zoom link: 307 Love, or https://fsu.zoom.us/j/98162962012 Record link: https://www.cs.fsu.edu/files/talks/2024_spring/Peizhong_Ju_Interview_Wed_0221.mp4 |
|
