From 1cf88ab9c4b4a25cd1bef7f76dae3c9a66f686b2 Mon Sep 17 00:00:00 2001 From: Sizzle Hsu Date: Wed, 13 May 2020 01:07:38 +0800 Subject: [PATCH] update README --- README.md | 17 ++++++++++------- 1 file changed, 10 insertions(+), 7 deletions(-) diff --git a/README.md b/README.md index 5267e2b..f04b837 100644 --- a/README.md +++ b/README.md @@ -2,8 +2,11 @@ ### Train an AI to crack the game! ### Live Demo -Launch and play the game (human player) -![demo](https://github.com/sizzle0121/2048-Game-and-AI/blob/master/img/demo1.gif) +Gameplay by AI +![demo](https://github.com/sizzle0121/2048-Game-and-AI/blob/master/img/demo.gif) + +Launch and play the game (human player) +![demo1](https://github.com/sizzle0121/2048-Game-and-AI/blob/master/img/demo1.gif) @@ -41,7 +44,7 @@ Here, I implement 6644-tuple network. ### Temporal Difference Learning -I implement TD(0) after state earning. +I implement TD(0) after state learning. The "after state" is like the Q(s, a) value. ![TDL](https://github.com/sizzle0121/2048-Game-and-AI/blob/master/img/TDL.png) @@ -49,7 +52,7 @@ The "after state" is like the Q(s, a) value. ### Future Development -> ->Add expectimax search to enhance the performance. ->Implement BitBoard to speed up training. ->Implementing DQN to extract features and train may be interesting as well. +> +>* Add expectimax search to enhance the performance. +>* Implement BitBoard to speed up training. +>* Implementing DQN to extract features and train may be interesting as well.