通俗演講

2019/01/10(Thu)     11:00 -12:30    1st Floor(auditorium)

Title

Deep Reinforcement Learning: Games to Industrial Applications

Speaker

Prof. I-Chen Wu

Department of Computer Science, NCTU

Abstract

In this talk, I will talk about one of the key AI technologies, development of Deep Reinforcement Learning (DRL), based on which Google’s DeepMind developed AlphaGo program that can defeat the Go champion which are thought to be not possible to happen within one or two decade. We will first review the latest development of Deep Reinforcement Learning (DRL), and then describe the DRL applications that we are developing. The applications include programs for games like AlphaGo (also including a new result for dynamic strength adjustment accepted by AAAI-19), bots for video games in industry, and random bin picking for robotics.

Google DeepMind發展的AlphaGo程式,打敗世界一流的李世石棋士、世界排名第一的柯潔,震撼全世界,掀起了全球AI的熱潮,開啟了AI的新時代。本演講將探討其中的關鍵性技術:深度強化式學習;談此技術如何應用到遊戲應用問題上,包括超越頂尖棋士的電腦圍棋,以及我們研發的動態棋力調整(將於AAAI-19發表),並應用之到終身學習圍棋的系統。然後,我們將更進一步探討如何深度強化式學習,應用於產業界,這裡將談及我們已經在研發的電玩遊戲產業相關應用以及機器人相關之應用。 (Poster)