Graph-Based Machine Learning and Neural Architecture Search

Thursday November 29, 2018
Location: Scaife Hall 214
Time: 3:00PM-4:00PM

This talk contains two parts. The first part is about semi-supervised learning via graphs, where we first introduce the concept of semi-supervised learning and then provide two graph-based methods from Google: Expander (label propagation via similarity graphs) and Neural Graph Machine (graph regularization for neural networks). The second part of the talk is about device-aware Neural Architecture Search (NAS). Neural Architecture Search (NAS) is famous for its effectiveness in searching for models that achieve state-of-the-art performances in a wide spectrum of applications, such as image classification and language modeling. In this talk, we provide an overview of commonly-used NAS techniques, and then propose a framework to search for neural architectures optimized for both device-imposed (e.g., power & inference time) and device-agnostic (e.g., accuracy) objectives.

Machine learner, software developer, and researcher: Da-Cheng Juan is a senior engineer at Google Research, exploring graph-based machine learning, deep learning and their real-world applications. Da-Cheng also holds the position of adjunct faculty in the Department of Computer Science, National Tsing Hua University. Previously, he received his Ph.D. from the Department of Electrical and Computer Engineering and his Master’s from the Machine Learning Department, both at Carnegie Mellon University. Da-Cheng has published more than 30 research papers in the related fields; in addition to research, he also enjoys algorithmic programming and has won several awards in major programming contests. Da-Cheng was the recipient of the 2012 Intel PhD Fellowship. His current research interests span across semi-supervised learning, convex optimization, deep learning, and energy-efficient computing.