Virtual Undergraduate Town Hall: Speeding up Deep Reinforcement Learning via Transfer and Multitask Learning
July 27, 2017
Virtual Undergraduate Town Hall
During the Virtual Undergraduate Town Hall Event, you will join students from around the world in a virtual mentoring event where you will learn about cutting edge research in the field of computing. You will have the opportunity to ask distinguished computer scientists any questions you might have.
Research Presentation: Speeding up Deep Reinforcement Learning via Transfer and Multitask Learning
Deep Reinforcement learning (DeepRL) has been increasingly popular as a way of learning to control agents without needing to hand-craft inputs. For example, DeepRL can learn to control video game characters directly from pixel-level input to maximize their scores! Unfortunately, DeepRL algorithms often suffer from high time and data complexity, which can be problematic for real-world applications. In this talk, Yunshu Du will review the basics of reinforcement learning and deep learning, and then describe some of the challenges associated with training a DeepRL agent. Lastly, Yunshu will present research on how to leverage two approaches, transfer learning (in which an agent can leverage learning in a previous task to speed up a new task) and multitask learning (in which an agent can learn to perform multiple tasks simultaneously). Results will be presented in the domain of Atari Games.
Mentoring Topic: How to Choose a Research Direction as an Undergraduate
While undergraduate research is rewarding, choosing the topic that suits you can be stressful. In this discussion, Yunshu will share her story of how she got involved in her current research interest, and will suggest what you should consider when looking for your first research project.
Post-Discussion Chat: Join Yunshu Du & Gail Murphy for a chat to continue the discussion about choosing a research direction, meet fellow students, and share your experiences.
Join us July 27th at 4:00pm ET
Speaker: Yunshu Du is a third year PhD student studying Computer Science in the School of Electrical Engineering and Computer Science at Washington State University. She has worked on research investigating Transfer and Multi-task for Deep Reinforcement Learning and Mining Student Data for Fitness. Yunshu attended the 2017 Grad Cohort.
Host: Gail Murphy, Professor in the Department of Computer Science and Associate Dean (Research & Graduate Studies) in the Faculty of Science at the University of British Columbia.
All are welcome to participate, register here.
Please join us 15 minutes before the presentation begins, in order to not miss any valuable information. The entire webinar event will last 1 hour. After the webinar we will host an interactive chat among attendees and the speaker, for 30 minutes. The event is broken into five sections:
- Yunshu Du presents on “Speeding up Deep Reinforcement Learning via Transfer and Multitask Learning”
- Open Q&A
- Yunshu Du discusses “How to Choose a Research Direction as an Undergraduate”
- Open Q&A
- Interactive chat forum with Yunshu Du and attendees
Are you part of an ACM-W chapter or Women in CS/CSE group? Check out our flyer on how to participate in this webinar event as a group!