TE
TechEcho
Home24h TopNewestBestAskShowJobs
GitHubTwitter
Home

TechEcho

A tech news platform built with Next.js, providing global tech news and discussions.

GitHubTwitter

Home

HomeNewestBestAskShowJobs

Resources

HackerNews APIOriginal HackerNewsNext.js

© 2025 TechEcho. All rights reserved.

Multi-Task Learning in Atari Video Games with Emergent Tangled Program Graphs

210 pointsby sengorkalmost 8 years ago

10 comments

bomdoalmost 8 years ago
I was a little surprised at the headline, since I expected &#x27;outperforms&#x27; to mean that it had better end-results, which is of course not the case. GP is just much faster due to it&#x27;s relative simplicity and the results are close enough to those achieved with NN and deep learning.<p>&gt; Finally, while generally matching the skill level of controllers from neuro-evolution&#x2F;deep learning, the genetic programming solutions evolved here are several orders of magnitude simpler, resulting in real-time operation at a fraction of the cost.<p>&gt; Moreover, TPG solutions are particularly elegant, thus supporting real-time operation without specialized hardware<p>This is the key takeaway and yet another reminder to not make deep learning the hammer for all your fuzzy problems.
评论 #14795404 未加载
smdzalmost 8 years ago
One 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&#x27;t a mathematician can analyze GPs with a lot of patience<p>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
评论 #14795119 未加载
评论 #14795630 未加载
评论 #14795265 未加载
nivwusquorumalmost 8 years ago
Those are really old results. They should compare to this one: <a href="https:&#x2F;&#x2F;arxiv.org&#x2F;pdf&#x2F;1511.06581.pdf" rel="nofollow">https:&#x2F;&#x2F;arxiv.org&#x2F;pdf&#x2F;1511.06581.pdf</a>
评论 #14794759 未加载
评论 #14794676 未加载
评论 #14795415 未加载
partycoderalmost 8 years ago
The convenient thing about Atari games is that there is usually a numerical score that can be used as input for the fitness function.
评论 #14795955 未加载
cshentonalmost 8 years ago
This is super cool, but it doesn&#x27;t outperform deep learning based RL methods.<p>In fact, I&#x27;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).
评论 #14796108 未加载
gouroualmost 8 years ago
Genetic Programming seems lightweight, what are some cool applications they have?
评论 #14797999 未加载
评论 #14796654 未加载
评论 #14796222 未加载
gouroualmost 8 years ago
What&#x27;s a good starting point for someone interested in building game AI?
评论 #14796775 未加载
评论 #14799624 未加载
评论 #14802252 未加载
nocoderalmost 8 years ago
This sounds interesting. I will like someone from the field of genetic programming on how this works and how it differs from current DL approaches.
评论 #14794882 未加载
评论 #14802670 未加载
jerianasmithalmost 8 years ago
I like GP, but the problem is AST. These can get huge. But the only advantage is ease of reverse engineering
99mistakesalmost 8 years ago
Slightly relevant, here&#x27;s a state of the art drone AI built using genetic fuzzy systems: <a href="https:&#x2F;&#x2F;www.forbes.com&#x2F;sites&#x2F;jvchamary&#x2F;2016&#x2F;06&#x2F;28&#x2F;ai-drone&#x2F;#50908d8b7081" rel="nofollow">https:&#x2F;&#x2F;www.forbes.com&#x2F;sites&#x2F;jvchamary&#x2F;2016&#x2F;06&#x2F;28&#x2F;ai-drone&#x2F;#...</a>