My entry for the Mario AI Championship 2012, implementing adaptive procedural level generation. Holds 1st place at the all-time points and percentage tally.
This approach works on some (dummy) states having the same name between the speeder and explorer schematics. Per cluster membership percentages (we really need a nickname for those, btw), jump between picking one transition or the other. See what happens.
This is the opposite, more complex, alternative to #58.
Quick way to do this: for each state we order the automata randomly according to odds. Then we try to get the dummy from each them until we get a success, and derive from there.
This approach works on some (dummy) states having the same name between the speeder and explorer schematics. Per cluster membership percentages (we really need a nickname for those, btw), jump between picking one transition or the other. See what happens.
This is the opposite, more complex, alternative to #58.