A series of notebooks experimenting with pymdp to create active inference agents.
These show emergent problem solving behaviour by trying to balance satisfying their preferences with gaining information about their environment (minimising their expected free energy).
Note that the numpy tic tac toe is very slow on a 4 * 4 board (5 mins per turn) but there is a JAX version in the devcontainer/src folder which is much faster, around 3 secs per turn.
