Research Topics

Deep Reinforcement Learning (DRL) integrates deep learning models into reinforcement learning framework to approximate action/state value or policy functions. Although it has been studied just for 3 years since 2013, the approach has produced outstanding results such as playing Atari video games at human level (Nature 2015) and beating the grand master of Go with AlphaGo (Nature 2016). Recently, there has been big demand on improving the deep reinforcement learning to tackle complex video games as an essential step towards real-world applications in robotics, conversational agents and recommendation systems. In this project, we propose to study cognitive models, deep architectures, and multimodal representation for advanced deep reinforcement learning.

A video game is an electronic game that involves human interaction with a user interface to generate visual feedback on a video device. In video games, content is often rather static and rigid, its script-based nature can lead to predictable and impersonal game play. On the other hands, adaptive video games attempt to model game players, predict future behaviors and provide tools for game contents generation. This project focuses on the use of computational intelligence algorithms with game logs, interaction data, and user evaluations for adaptative video games.