Talk:Neuroevolution of augmenting topologies

Untitled
Why to hang on to this page: The other "dupe" page is the one with the wrong name. For years that page has carried the mistaken title "NeuroEvolution of Augmented Topologies," apparently because someone made a typo long ago when they first created the page. Finally someone appears to be trying to fix this error with this new page, which has the correct name "NeuroEvolution of Augmenting Topologies." Note the difference is AugmentED (wrong) vs. AugmentING (right). So THIS page on NEAT should be kept and the OTHER one should be deleted. This is easy to verify: The original author of NEAT has many papers written on the subject at his site http://www.cs.ucf.edu/~kstanley. You can verify there that they all say "NeuroEvolution of Augmenting Topologies"

132.170.16.56 04:59, 12 April 2007 (UTC)


 * When a page needs to be renamed it is moved to the new name, not cut and pasted under a newly created article. Doing it by cutting and pasting destroys the page history. I checked the external links, confirmed to my satisfaction that it should be augmenting, and moved it. The other page has been deleted.--Fuhghettaboutit 19:55, 14 April 2007 (UTC)


 * I did the cut and paste. Sorry I'm a wiki newbie. I'll do a move next time. Thanks for fixing it though Fuhghettaboutit. =) JockoJonson 20:07, 8 May 2007 (UTC)

Wrong title
The title of this article is wrong, it should be spelled with capitals: NeuroEvolution of Augmenting Topologies. See for example Ken Stanley's website. - Simeon (talk) 17:23, 4 May 2009 (UTC)


 * It used to be all caps but some someone changed it. Apparently the single cap word with the rest lower case is standard procedure? Not sure ... —Preceding unsigned comment added by 12.8.194.30 (talk) 13:37, 7 July 2010 (UTC)

Timers
"The basic idea is to put the population under constant evaluation with a "lifetime" timer on each individual in the population. When a network's timer expires its current fitness measure is examined to see whether it falls near the bottom of the population, and if so it is discarded and replaced by a new network bred from two high-fitness parents. A timer is set for the new network and it is placed in the population to participate in the ongoing evaluations."

It isn't clear whether the timers on "each individual" are different from "a network's timer", or what is meant by "individual in the population". Are the "individuals" distinct instances of networks, or neurons, or connections, or weights, or what? In other words, what does the "population" consist of? I assume that "it" refers to "network" (rather than "network's timer"), and that the "fitness measure" is that of the network as a whole, but that, too, isn't clear. The new network is "bred" from "parents", a concept which hasn't so far been introduced. Does the last "it" refer to the new network (rather than the timer which is set for it)? The entire paradigm expressed here does not seem to follow naturally from the preceding discussion. Unfree (talk) 19:13, 11 December 2009 (UTC)

Introducing HyperNEAT
"HyperNEAT is specialized to..." doesn't address what HyperNEAT is. I get the impression that it is an extension of NEAT to multiple dimensions; true? What is the nature of a "dimension", or of geometry in general, in this context? Unfree (talk) 19:29, 11 December 2009 (UTC)

What about WANN (Weight Agnostic Neural Networks)?
https://weightagnostic.github.io — Preceding unsigned comment added by 202.75.86.157 (talk) 06:06, 20 January 2020 (UTC)