Wei-Yu Chen | 陳威侑

I'm Designer

About

Hi, I am Wei-Yu Chen. I received B.S. degree in Electrical Engineering and B.A. degree in Economics from NTU. Before that, I received my high school diploma in science class from National Tainan First Senior High School. I am interested in reinforcement learning, machine learning, and economics. Hoping to make a contribution to the world, I will try my best to learn, research, and explore.

Experiences

The only source of knowledge is experience. by Albert Einstein.

Education

B.S. in Electrical Engineering

B.A. in Economics

Sep 2020 - Jun 2024

National Taiwan University, Taipei, Taiwan

  • Major in Electrical Engineering and double major in Economics.
  • Top 4% of the class.

    • High School Diploma

    Sep 2017 - Jun 2020

    National Tainan First Senior High School, Tainan, Taiwan

  • Science Class (Program for students gifted in science)

    • Professional Experience

    Research Assistant

    Feb 2023 - Jun 2024

    Institute of Information Science, Academia Sinica, Taipei, Taiwan

    • Enhanced AlphaZero's head-to-head win rate on Go and Othello by 15% through strategic selection of intermediate states as self-play starting positions.

    Undergraduate Research Assistant/Intern

    Jul 2023 - Aug 2024

    Institute of Information Science, Academia Sinica, Taipei, Taiwan

    • Reduced decoding MSE loss on Atari games by 30%, improving interpretability of MuZero's hidden states.

    Undergraduate Research Assistant

    Sep 2022 - May 2024

    NTU Robot Learning Lab, Taipei, Taiwan

    • Leveraged large language models and semantic embeddings to identify relationships among predefined MuJoCo skills, enabling a 10% increase in new skill training speed by transferring knowledge from existing skills.

    Summer Research Assistant

    Jun 2023 - Sep 2023

    Hybrid Robotics, UC Berkeley, California, USA

    • Developed an emotional and meaningful script generation system for the Unitree A1 quadruped robot using dual OpenAI API instances for self-adjustment.
    • Predefined a set of skills, each with a Blender-rendered video, composing scripts from sequences of these skills.

    Publications

    *Equal Contribution

    Interpreting the Learned Model in MuZero Planning
    Hung Guei, Yan-Ru Ju*, Wei-Yu Chen*, and Ti-Rong Wu
    The 29th International Conference on Technologies and Applications of Artificial Intelligence (TAAI 2024)
    Paper / Website / Code

    Incorporate a decoder network to endow MuZero interpretability.

    Awards

    TAAI 2024 Runner Up Paper Award
    MakeNTU AUO Gold Sponsor Prize

    Contact

    If you are interested in my research or experiences, please do not hesitate to contact me.

    Phone

    +886 978 515 908

    Email

    jasperorachen@gmail.com