snake-ai
snake-ai copied to clipboard
add a circle detection as penalty to agent and add mps support to the MLP
在训练过程中对蛇增加了一个惩罚。如果蛇的身子包住了空格子,将会根据包围住的格子大小受到惩罚。该策略目的在于使蛇在一定程度上折叠自己的身子。(图片显示在某一局蛇的长度达到了103)
A penalty was added to the snake during training. If the snake's body wrapped around empty cells, it would be penalized according to the size of the cells. The purpose of the strategy is to make the snake fold its body to some extent. (Picture shows a snake reaching 103 in one game)