Bei Peng
office: Dana Hall 3
I am a PhD student who works with Matthew E. Taylor in School of Electrical Engineering and Computer Science in Washington State University.

My research mainly focuses on interactive machine learning, where the main goal is to combine AI systems with human intelligence, such that the systems can learn more efficiently from humans when needed, and non-expert humans can access the benefits of machine learning. This is important since the deployment of real-world robotic systems requires to be adaptive to many different complex human environments. A reliable AI system needs to work with humans, not alone. We need to better understand how do human teachers want to teach the agent so that we can better model them to gain more effective learning interaction. So we have been working on exploring how to better interpret, incorporate, and allocate human feedback to speed up agent learning. We also work on designing a better representation of the learning agent to elicit a more natural and effective learning interaction between the human trainer and the learner. Currently, we focus on curriculum learning in the context of reinforcement learning.