Machine learning on a simple grid game by RoastBeef#11541
This code is over 6 months old. The code may have expired and might no longer function.
A simple grid game where the agent is trying to look for a purpose in its artificial life. It doesn't know he has a goal or enemies yet from the start. It goes around its environment and depending on where it lands, it will either get a reward or get killed by its enemy. It learns through reinforcement training, just like how you would give a dog a treat for something it did well. This is just a proof concept.
Data can be stored and reused even after you restart the game. Execute any voice line, wait for it to compile. In the workshop inspector, in the global tab, click the button that looks like (X). Once you've copied, go into the workshop, make a new global rule and paste it in. It saves your training data and the position of the enemies. Epsilon will be reset but will easily go down.
This post was originally hosted on workshop.elohell.gg, a site which is no longer online. If this was originally your post, you may click the button below to transfer this post to your Workshop.codes account or delete this post entirely.Archive Actions