From 16c0913a3b259f683a0eac2873514028ba7cd5dd Mon Sep 17 00:00:00 2001 From: Dominik Roth Date: Fri, 13 May 2022 17:06:35 +0200 Subject: [PATCH] Added update about current state to README --- README.md | 5 ++++- 1 file changed, 4 insertions(+), 1 deletion(-) diff --git a/README.md b/README.md index 23784dc..7550d32 100644 --- a/README.md +++ b/README.md @@ -19,4 +19,7 @@ A new AI is created by subclassing the State-class and defining the following fu ### Current state of the project The only thing that currently works is the AI for Ultimate TicTacToe. It uses a trained neural heuristic (neuristic) -You can train it or play against it (will also train it) using 'python ultimatetictactoe.py' +You can train it or play against it (will also train it) using 'python ultimatetictactoe.py' + +The performance of the trained neuristic is pretty bad. I have some ideas on what could be the problems but no time to implement fixes. +(Focus on the ending of games at the beginning of training; more consistent exploration-depth for expanding while training; ...)