A philosophical tangent: Evolution did not always produce the best outcomes. Would neuroevolution be vulnerable to similar effects? Probably a good research area. https://www.wired.com/2009/07/st-best-5/
What’s the definition of best
Evolution Is the New Deep Learning
11–20 of 242 posts
Re: Evolution Is the New Deep Learning
#12A philosophical tangent: Evolution did not always produce the best outcomes. Would neuroevolution be vulnerable to similar effects? Probably a good research area. https://www.wired.com/2009/07/st-best-5/
A can think of a few objections to that argument. First, evolution isn't 'over' for any of the species on that list, and if any of those seemingly maladaptive traits really do cause serious problems for those species, we would expect them to be selected against (there are some cases where they are associated with other traits that are important and helpful, and so won't be tossed overboard even then). Evolution also proceeds fairly slowly, in the sense that it may take many generations for a mutation to really become a positive adaptation and not just an oddity.
Re: Evolution Is the New Deep Learning
#13Re: Evolution Is the New Deep Learning
#14As expected, the article seems to be a typical content marketing piece. If you're looking for real insights into evolutionary algorithms, specifically "neuroevolution", I highly recommend to read this article: https://www.oreilly.com/ideas/neuroevolution-a-different-kin... I enjoyed it much more than - what feels like - a quickly thrown together marketing piece with no real value for the reader.
Thank you, saved it for later. Do you have any other links to offer?
Here's a recent survey / observational science paper by some prominent "neuroevolution" / A-Life researchers.
https://arxiv.org/abs/1803.03453. I found this refreshing because it's rare that science papers talk about the debugging and experimental process and debugging journeys underlying this research.
Re: Evolution Is the New Deep Learning
#15Having studied this extensively back when they were called Genetic Algorithms, I would like to offer a few insights.
1) One of the biggest reasons they fell out of favor for more "mathematical" approaches was that no one could really explain why exactly they worked. It makes sense on the surface that "survival of the fittest" and doing something akin to multiple stochastic gradient descents would work, but no one has really been able to produce a mathematical proof as to why.
Since other folks are producing good examples of "explainable AI", I don't know how Genetic Algorithms/programming could be made 'explainable' as to why they achieved an optimal solution other than hand-waving to how evolution works in nature.
2) The most important thing to define is the fitness function, this defines what the search space looks like and how easily a globally optimal solution can be derived. For a good example of an interesting search space that a genetic program would have a difficult time with, see Schwefel functions [0]. Back when I researched these things closely, my intuition was that reality rarely fits neatly into good fitness functions and I felt that at the point you are understanding the problem, you may just be better off with a direct approach, which leads to
3) Genetic programming should only really be considered when there are no known alternatives or they are way too computationally expensive.
In either case, I would welcome a resurgence in a topic I once knew quite well, though I haven't been in that field for a few years now.
[0] https://jamesmccaffrey.files.wordpress.com/2011/12/schwefels...
Re: Evolution Is the New Deep Learning
#16As expected, the article seems to be a typical content marketing piece. If you're looking for real insights into evolutionary algorithms, specifically "neuroevolution", I highly recommend to read this article: https://www.oreilly.com/ideas/neuroevolution-a-different-kin... I enjoyed it much more than - what feels like - a quickly thrown together marketing piece with no real value for the reader.
Not quite a typical piece - it managed to make it to #2 on HN.
I upvote links that promote discussion of interesting concepts, even if the article itself is really bad - like this one.
I couldn't bother reading the original article, but designing neural networks via evolutionary algorithms is a very interesting concept I wasn't aware of.
Re: Evolution Is the New Deep Learning
#17The problem is in the genotype-phenotype map and getting that right. We should be looking at developmental evolutionary neural networks. That is, evolutionary algorithms evolving the genetic architecture (genotype) that generates (development) neural networks (phenotype).
Re: Evolution Is the New Deep Learning
#18A philosophical tangent: Evolution did not always produce the best outcomes. Would neuroevolution be vulnerable to similar effects? Probably a good research area. https://www.wired.com/2009/07/st-best-5/
Yeah, agreed -- evolution doesn't produce best, just good enough. But for large classes of 'problems', we can't (from a computational complexity standpoint) find best outcomes efficiently.
'Evolutionary' techniques, are, in some sense, 'trial and error'. For many tasks where we already have efficient algorithms, this isn't particularly useful. However, where we DONT have efficient algorithms, or know what the concept of what an 'efficient' algorithm even is, trial and error techniques, like evolutionary algorithms, seem like great choices.
Re: Evolution Is the New Deep Learning
#19Having studied this extensively back when they were called Genetic Algorithms, I would like to offer a few insights. 1) One of the biggest reasons they fell out of favor for more "mathematical" approaches was that no one could really explain why exactly they worked. It makes sense on the surface that "survival of the fittest" and doing something akin to multiple stochastic gradient descents would work, but no one has…
Wait, I'm confused...you're saying genetic approaches fell out of favor because they're basically just stochastic gradient descent? Most of modern DL relies heavily on SGD at various points during training.
My impression is that they fell out of favor precisely because don't actually use any gradients, and end up converging on good maxima slower than you could if you used the gradients from the net. Am I off base here?
Re: Evolution Is the New Deep Learning
#20Having studied this extensively back when they were called Genetic Algorithms, I would like to offer a few insights. 1) One of the biggest reasons they fell out of favor for more "mathematical" approaches was that no one could really explain why exactly they worked. It makes sense on the surface that "survival of the fittest" and doing something akin to multiple stochastic gradient descents would work, but no one has…
Regarding 1, that seems surprising: biology has had many mathematical proofs and models (for almost a century now, e.g. Fisher) explaining why evolution works. Evolution works in nature for non-hand-wavey reasons, doesn't the same logic justify artificial evolution in genetic algorithms?