[CCoE Notice] PhD Dissertation Defense Detail - Kuo Chun Tsai

[Knudsen, Rachel W]

PhD Dissertation Defense Detail - Kuo Chun Tsai

Name: Kuo Chun Tsai
Advisor: Dr. Zhu Han
Date: 4/20 at 10:00 am
Zoom link: https://zoom.us/j/835186397<https://www.google.com/url?q=https://zoom.us/j/835186397&sa=D&usd=2&usg=AOvVaw2nH0rA7ZEhvoOol3P0UWFL>
Title: Machine Learning From Application to Theorem
Abstract:

With the rapid improvement of technology in the past few decades, many daily life tasks have been replaced by various machines. With the rise of the machine learning technique, machines became faster and smarter than humans in certain scenes. Machine learning is the study of computer algorithms that provides systems the ability to automatically learn and improve from experience without being explicitly programmed. In this dissertation, we try to advance machine learning knowledge in both applications and theories with the following three studies.

We first study the potential of machine learning by applying this technique into the mobile social networks edge caching problem. Low data transmission latency is one of the top priorities in the system. We found out that the machine learning model with long short-term memory (LSTM) structure can determine the relevance of the data to the user. Based on the machine-learned preferences of mobile users, we match the cached data with the base stations. In the simulation, we can see that by caching the relevant data close to the interested user, the data transmission latency within the network has highly reduced. With the success of connecting the machine learning technique with the edge caching and computing problem, we further bring machine learning into a real-world problem.

In the oil and gas industry, the first arrival picking in seismic imagining is performed to determine the underground structure. This process highly depends on human labor where expensive experts need to manually look at the data one by one, which is too much time consuming and inconsistent judgment. With the help of machine learning with the semi-supervised learning and transform learning technique, we not only reduced the processing time but also produced consistent results by learning from different experts' experiences.

To link the machine learning, we bring the machine learning technique into game theory. Game theory is a very profound study on distributed decision-making behavior and has been extensively developed by many scholars. However, many existing works rely on certain strict assumptions that might not be practical. With the robust data mining and learning ability of the machine learning model, we can relax the assumptions and study the opponent's behaviors under the condition of limited information given. Our proposed deep reinforcement learning learns why and how the rational opponent plays, instead of just learning the regions for corresponding strategies and actions.


-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://Bug.EGR.UH.EDU/pipermail/engi-dist/attachments/20200413/28ccd9fd/attachment.html