[CS] Zhengxu Xia Dissertation Defense/Sep 27, 2024

[via cs]

This is an announcement of Zhengxu Xia's Dissertation Defense.
===============================================
Candidate: Zhengxu Xia

Date: Friday, September 27, 2024

Time:  10:30am -11:30am CT


Location: JCL 236

Title: Guided optimization for learning-based solutions for networking

Abstract: Unlike machine learning (ML) models trained for applications like object detection and classification which consume data and yield the output directly to the end users, many ML models trained for networking problems are often embedded between the application layer and the network due to the layered characteristics in computer networks. For example, reinforcement learning (RL) models trained for congestion control and adaptive bitrate streaming sit between the network and the video codec. Therefore, the diversity and dynamics of the application layer and the network can both affect the performance and the behavior of a ML model trained for networking problems. This work introduces how to guide the ML model training for networking to adapt to diverse network environments and application layer changes by leveraging heuristic based solutions or application layer entities like video codecs.


Advisors: Junchen Jiang

Committee Members: Junchen Jiang, Nick Feamster, and Francis Y. Yan