Live data from Hacker News

Watch cars evolve using genetic algorithm

rednuht.org

71–76 of 76 posts

Re: Watch cars evolve using genetic algorithm

#71
post #15
post #11

I wrote the original version of this nearly 20 years ago. Cool to see it still pop up here from time to time. Still runs in the browser thanks to Ruffle: https://peteshadbolt.co.uk/posts/ga/

On Firefox here, the "Save Local Population" option seems to crash. Any idea why that might be happening? (Amazing site btw - every time it pops up I end up spending much too long with it).

Where do you see the "Save Local Population" option? When you say it "seems to crash", do you mean Firefox or the Flash animation?

Re: Watch cars evolve using genetic algorithm

#72
post #63
post #33

Earlier quoted context omitted.

Sounds like you're running with a single elite clone. That's a really bad idea in genetic heuristics for exactly the phenomenon you mention.

Is the correct number 0 or >1?

Note my use of the term "heuristic" and not "algorithm." There is not a correct number. The correct approach is to fiddle with the parameters; record good populations; occasionally restart from scratch... we call this "hyperparameter tuning" to make ourselves feel better about the process.

Re: Watch cars evolve using genetic algorithm

#73
post #11

I wrote the original version of this nearly 20 years ago. Cool to see it still pop up here from time to time. Still runs in the browser thanks to Ruffle: https://peteshadbolt.co.uk/posts/ga/

As internet distractions go, the original was one of the most memorable I ever came across. A friend and I used to leave them running over lunch and see who was winning when we got back. Thanks for the fun :-)

Re: Watch cars evolve using genetic algorithm

#74
post #11

I wrote the original version of this nearly 20 years ago. Cool to see it still pop up here from time to time. Still runs in the browser thanks to Ruffle: https://peteshadbolt.co.uk/posts/ga/

Thanks for writing this all those years ago. It made for a fun demonstration when I was learning genetic algos in college.

Re: Watch cars evolve using genetic algorithm

#75
post #65

might go much faster if it recorded a set of states where ancestors died shortly after and then test all the new candidates against those states. The new candidate might actually survive because its prior history kept it from ever getting into that particular death state, but I think biasing towards designs that don't immediately die in those hard cases is good anyways, since given a long enough run it would likely e…

Recessive genes, essentially.
Post reply on HN