Hi I'm Laige. I am now a master student in the University of Tokyo. My research theme is deep reinforcement learning in the game of Minecraft. I focus on agent control, policy decision and sample efficiency.
I like animations, games, Taichi Mukai, RADWIMPS and SEKAI NO OWARI.
Propose a new method to combine Proximal policy optimization algorithms with action branching architecture, conduct experiments in several different environments and utilize human demonstration data.
Results: proposed method accelerates the learning process and achieves higher scores than baseline. Research paper accepted by the 24th Game Programming Workshop as oral presentation.
Game AI development to finish specific tasks in Minecraft: Utilized life-long learning conception in a hierarchical Deep Q-Network architecture in Python with Pytorch.
Published a paper "Learning basic skills to survive the first day in Minecraft with life-long learning" in the 23th Game Programming Workshop in Japan.
Keep your passwords safe! Design, develop and release a decentralized password management tool by using blockchain with team members.
Computer skills/Other
The School of Information Science and Technology, The University of Tokyo. Apr. 2018 - Present.
Master of Information and Communication Engineering.
Zhejiang University, Sep 2013 - Jun 2017.
Hackathon participant: Design, develop and release a decentralized password management tool by using blockchain with team members. Won Technology Completion Special Award.
Conducted survey about abstractive text summarization algorithms. Made an interactive application to summarize input texts by applying pre-trained models.
Hydraulic pump design engineer assistant: designed components and managed drawing catalog.
Standardized documents for pump design team.
In a welfare organization for disabled people named Dual Blessing Bhd, made handicrafts together with disabled people, organized charity bazaar and edited promotional video.
The University of Tokyo, Hongo Campus