Genetic Programming seems lightweight, what are some cool applications they have?
Multi-Task Learning in Atari Video Games with Emergent Tangled Program Graphs
41–50 of 58 posts
Re: Multi-Task Learning in Atari Video Games with Emergent Tangled Program Graphs
#42This is super cool, but it doesn't outperform deep learning based RL methods. In fact, I'm not sure how much more compute efficient than something like A3C it would be. That can produce 4x the score of DQN in a comparable number of hours (and on a CPU).
A3C is only ever run on one game at a time[0]. This paper gets good performance on all games with the same agent
[0] read as: I have only seen papers with 1 agent per game for A3C
Re: Multi-Task Learning in Atari Video Games with Emergent Tangled Program Graphs
#43What's a good starting point for someone interested in building game AI?
Re: Multi-Task Learning in Atari Video Games with Emergent Tangled Program Graphs
#44This is super cool, but it doesn't outperform deep learning based RL methods. In fact, I'm not sure how much more compute efficient than something like A3C it would be. That can produce 4x the score of DQN in a comparable number of hours (and on a CPU).
A3C is only ever run on one game at a time[0]. This paper gets good performance on all games with the same agent [0] read as: I have only seen papers with 1 agent per game for A3C
So it can train on one game and play without training on a previously unseen (but also atari) game? That's pretty neat, DQN and A3C certainly can't do that.
Re: Multi-Task Learning in Atari Video Games with Emergent Tangled Program Graphs
#45I like GP, but the problem is AST. These can get huge. But the only advantage is ease of reverse engineering
Re: Multi-Task Learning in Atari Video Games with Emergent Tangled Program Graphs
#46Genetic Programming seems lightweight, what are some cool applications they have?
It's just a toy, but it's pretty entertaining to watch:
Unfortunately, it requires Flash.
Re: Multi-Task Learning in Atari Video Games with Emergent Tangled Program Graphs
#47Earlier quoted context omitted.
What is "exponential growth"!
It's more logarithmic, isn't it? Right now we're at the beginning stage, where there's massive discoveries and changes happening, but in say, 50 years time, there won't be much changing year-to-year.
Logarithmic or logistic?
Re: Multi-Task Learning in Atari Video Games with Emergent Tangled Program Graphs
#48Genetic Programming seems lightweight, what are some cool applications they have?
A couple of years ago I read about some printer manufacturer evolving the shapes of their nozzles. Apparently the problem was nontrivial and rather than doing complicated analyses up front they found it more efficient to simply generate and fabricate random permutations of shapes and evolve them over many generations. The results were better than any human designed ones, apparently.
Re: Multi-Task Learning in Atari Video Games with Emergent Tangled Program Graphs
#49What's a good starting point for someone interested in building game AI?
I'm far from knowledgable in the field, but I did a survey some time ago and think these items should provide a decent basis:
Dijkstra and A* Pathfinding, Finite State Machines, Decision Trees, Hierarchical Task Networks (SHOP, etc)
Keep in mind that game ai algorithms are all about decision taking, there's little "intelligence" involved, unlike the broader aim of "general" ai.
Re: Multi-Task Learning in Atari Video Games with Emergent Tangled Program Graphs
#50One of the huge benefits of GPs over NNs is the ease of reverse engineering a GP tree compared to NN models. Its not effortless however. Its just not mathematically complex like NNs i.e. a programmer who isn't a mathematician can analyze GPs with a lot of patience EDIT: I have found GPs to be relatively slow-to-very-slow. But very likely that is because of the lack of interest and development compared to NNs
Would some form of computer-aided tool be able to help the programmer analyze GPs? It sounds like there's some repetitive process when you say 'GPs with a lot of patience'. I was thinking along the line that if such a tool is possible, then it might be possible to let the development and evolution of GPs iterate faster than deep learning. I found this thread really interesting, because of the small size, such impleme…
I quit working at shoprite and now I make $35h - $80h...how? I'm working online! My work didn't exactly make me happy so I decided to take a chance on something new…after 4 years it was so hard to quit my day job but now I couldn't be happier.
........... http://cutt.us/7N2sG