schrum2 / MM-NEAT

Modular Multiobjective (Hyper) Neuro-Evolution of Augmenting Topologies + MAP-Elites: Java code for evolving intelligent agents in Ms. Pac-Man, Tetris, and more, as well as code for Procedural Content Generation in Mario, Zelda, Minecraft, and more!
http://people.southwestern.edu/~schrum2/re/mm-neat.php
Other
50 stars 20 forks source link

Implement shapeIsWorthSaving for the Missile Fitness and Cannon Fitness #914

Closed schrum2 closed 1 year ago

schrum2 commented 1 year ago

See #913 to see what this is about. A simple implementation of this will give you a way to save some evolved canons to help with #906

schrum2 commented 1 year ago

Implement this and base it directly on fitness. Say that if 90% of the target is destroyed, then save

TjRaffert commented 1 year ago

We are getting kicked out of the lab soon so I will test when we have enough time.

schrum2 commented 1 year ago

Lost track of this because of how things ended. What is the state of this issue now?

TjRaffert commented 1 year ago

I am running it with the 90% and I confirmed it works when I manually blow it up. It ran all night and didn't save any shapes. I may rerun it with the setting turned down to 60% so we can look at some results.

schrum2 commented 1 year ago

If you complete #911 then we can get results at the end of evolution even if they don't earn high fitness, though this may take too long to be realistic. Still, it is another issue that is important

schrum2 commented 1 year ago

Make the percentage be a command line parameter

TjRaffert commented 1 year ago

Somehow the line in parameters got deleted but I added it back. I had already tested it and it works. You have to adjust it based on the test you are running because some target orientations are much harder than others. I got some video of some of the results.