KaigeLLM + MCTSA proper agent ought to have solid planning and learning ability through trial-and-error rather than only memory and in-context learning…Aug 18Aug 18
KaigeGame Level Difficulty EstimationThis article reviews papers on AI estimating game level difficulty methods in the literature.Aug 8Aug 8
KaigePredict Game Outcome before Game-Playing StartsThis paper propose a method to predict the game outcome of two teams each consists of 5 players.Aug 5Aug 5
KaigeRL Algorithm Distillation with Casual TransformerThis article goes through the following paperAug 1Aug 1
KaigePPO-V3: PPO+ Tricks of Dreamer-V3This paper combines tricks proposed in DremaerV3 with PPO.Jul 30Jul 30
KaigeDrQV2, CURL and DSAC-TLets’s go through the details of DrQ-V2, CURL and DASC-T algorithmsJul 28Jul 28
KaigeDreamerV3 and MuzeroBoth DreamerV3 and Muzero are model-based RL algorithms. This article dives deep into the details trying to understand these algorithms and…Jul 26Jul 26
KaigeUse dozens GPUs and hundreds CPUs to train RL on AWS SagemakerThis article shows how to train RL with dozens and hundreds CPUs on AWS SageMaker.Jul 22Jul 22
KaigeHow to write custom callback in Ray RLlib?Suppose we try to train RL NPC for a game which contains many level. The game starts from a random level eat each new episode (reset). This…Jul 21Jul 21