
We (@yannakakis, Ruck Thawonmas and I) ran a seminar at Shonan in Japan on Artificial General Intelligence in Games, with 32 researchers from industry and academia participating. Here's the report: shonan.nii.ac.jp/seminars/130/
Reinforcement Learning
79 posts

@GameDevPipeline
@METU_ODTU MMI 706 Course / 2018-2019 Spring Semester - by Asst. Prof. Elif Surer (@ElifSurerMETU)

We (@yannakakis, Ruck Thawonmas and I) ran a seminar at Shonan in Japan on Artificial General Intelligence in Games, with 32 researchers from industry and academia participating. Here's the report: shonan.nii.ac.jp/seminars/130/




DeepMind releases "dm_control" a suite of continuous control reinforcement learning environments, based on the proprietary MuJoCo framework. github: github.com/deepmind/dm_co… paper: arxiv.org/abs/1801.00690











Initial details on #AlphaStar. As expected, bootstrapped from imitation learning, and then using population-based self-play (Lamarckian evolution with RL). Finally, there's a specialisation to find a set of the least exploitable agents (Nash equilibria).






