Structured reinforcement learning
Jungseul Ok is an assistant professor in the Department of Computer Science and Engineering, and a member of Machine Learning Lab at POSTECH. He completed Ph.D program in School of Electrical Engineering at Korea Advanced Institute of Science and Technology (KAIST), South Korea, under the supervision of Prof. Yung Yi and Prof. Jinwoo Shin. After graduation, he worked with Prof. Alexandre Proutiere and Prof. Sewoong Oh as a postdoctoral researcher, respectively, in School of Electrical Engineering at KTH, Stockholm, Sweden, and Paul G. Allen School of Computer Science & Engineering, University of Washington, WA, US. His research focus is theoretical machine learning on a wide spectrum of topics, including neural network analysis, Markov decision process, structured reinforcement learning, recommender system, wireless network, social network, crowdsourcing system, and influence maximization.
In the last few years, we have seen impressive advances in machine learning: image recognition, self-driving car, language translation. Such advances are thanks to complex deep network models which can learn huge dataset. However, to see fully self-driving cars or human-level translator, we need much larger models and thus larger dataset, while the data supply is limited by budget and time constraints. Hence, high data efficiency is one of main research stream in machine learning, in which it is important not only to adaptively collect data for most learning gain, but also to fully utilize the observed data as much as we can. In this talk, I will be introducing how to develop high data-efficient system in the framework of structured reinforcement learning, where we study optimal use and collection of data when we know a prior knowledge such as physical law or meta information.