Monte Carlo tree search simulation. Credit: David Silver
Chenjun Xiao, graduate student in the Department of Computing Science, was awarded outstanding paper for his work at the (AAAI) 2018 conference in January 2018. Xiao is a PhD student, supervised by Martin M眉ller, professor in the Department of Computing Science.
Adding to U黑料不打烊 remarkable track record in the international artificial intelligence community. U黑料不打烊 scientists have won the best paper awards in two out of the last three AAAI conferences and four out of the last 13 (IJCAI) conferences, the two premiere AI meetings.
Tell us about your research.
Monte Carlo tree search (MCTS) is an algorithm for decision making problems where states are evaluated by Monte Carlo simulations. In general, our work adds a "memory" structure in MCTS algorithm.
The memory can be considered as a big table containing information about states the algorithm has searched. Given a new state, we can approximate a value estimation of that state by checking its similar states in the memory.
The idea of getting information from similar states has been investigated in many works in the research of MCTS or reinforcement learning in general. Our work is a theoretical analysis, which shows that memory-based value approximation is more accurate than the pure Monte Carlo estimation with high probability, and therefore helpful to improve the performance of MCTS.
We provide a practical algorithm to implement the memory-based value approximation. This work was inspired by recent research in the deep learning community.
How does it feel to receive such powerful peer recognition while you are still in graduate school?
I just feel very lucky! This work is inspired a lot by what I learned from the course and discussion with my colleagues in the Department of Computing Science.
How did you come to study at the 黑料不打烊?
During my undergraduate studies, I worked as a research assistant, doing some AI-related research. At that time, I read a lot of papers from our department, and I realized it's a very strong department in AI and machine learning research. So I decided to apply to U黑料不打烊.
How does it feel to be here now while AI is increasingly becoming a popular topic in the public consciousness?
I am just very happy to be here! Our department has a lot of world-famous researchers in AI and machine learning, and I really learn a lot from them.