... This is somewhat of a mechanistic model. The 'wave' in question is the 'flowing neurochemical signal' that travels through a biological neural network. ...
It's been just a few months since I've started studying NEAT and I'm sorry if my question is very basic or has already been asked a lot of times before (I've...
My company is in the process of developing a NEAT accelerator. This would be a combination of hardware and software that accelerates the process of evolving a...
Hi Shane, all, I think for many of us your email raises quite a few questions! Firstly the web page at the domain you gave states "Analysis for the financial...
About a month ago someone on the Numenta forums posted an idea about modifying NEAT to optimize architectures for HTM (hierarchical temporal memory) systems,...
Hi Carolina, Basically you can either keep the list of innovation numbers for the current generation only or keep it forever. Try to experiment with each and...
Hi Petar, I'm the guy Carolina mentioned as a friend on her message. We're aware of NEAT's FAQ and some old topics on that matter. But basically our point is...
Shane, Naturally I'm very interested in what a NEAT accelerator actually does. My research group certainly does a lot of work with NEAT so we would actually...
... temporal ... Derek, thanks for pointing that out; it's an interesting suggestion. HyperNEAT might also be applicable to HTM's, which I'm guessing would ...
Hi Cesar, I think this is possible, but the genome crossover/comparison procedures may take longer be harder to implement. First you will have to search for...
... With ASICs, I think the gain could be considerably more, if, as you said, the hardware actually speeds up the most expensive part of NEAT. ... I agree.....
NEAT uses speciation, according to Ken, for two reasons: protecting innovation and "preventing incompatible genes from crossing over" (from his dissertation)....
Derek, I agree that it's interesting to question the mechanism and even the justification behind "protecting innovation." Certainly such declarations can...
... I certainly see the justification behind protecting innovation. It's the particular method of subdividing the population as a way of doing it that I'm...
... Derek, I'd have to review the code to be sure, but I think our NEAT implementation differs from Ken's in that we let parents survive unmutated from one...
... Hi guys, just to interject on this point. Intuitively: No, a pig and a bird are much more dissimilar from (for example) two strategies for balancing a ...
The program use higher fitness value to mark better individuals, but I am working on a problem which requires minimum optimization while using errors as...
... I wonder if, when creating NEAT, Ken tried using demes? Where, each deme has four neighbors, and migration (random or best) happens occasionally between...
... Have you tried using root mean square error[1] as your fitness? [1]: http://en.wikipedia.org/wiki/Mean_squared_error -- Stephen Waits steve@... ...
... doing ... each ... (as ... but ... I never tried what you call demes, or any kind of geographic-based mating structure. I am guessing that it would not...
For what it's worth, here's my $0.02.... Neat works on two quite different time-scales. You have the 'optimising weights' timescale which is fast, and the...
Hello sorry about the delay in responding. The NEAT accelerator speeds up the most time consuming portions of the NEAT algorithm only. Unfortunately it won't...
... Thanks very much for your advice, but if I use root mean square error, doesn't that mean the best individual gets the lowest fitness value? Since I am new...
... Thanks very much for your advice, but if I use root mean square error, doesn't that mean the best individual has the lowest fitness value? Since I am new...
... Yes, see below. ... For that you'll need to change the activation function to something that goes from say [-1,1]. You could transform a [0,1] activation ...
If you have raw fitness (rf) being the RMSE, then you can let the actual fitness (f) be... f=1/(1+rf) ... which will give you values of fitness in (0,1) with...
I have an input pattern and an output pattern both with random data set, my goal then is to form an ANN which can reponse to the input pattern by yielding the...
Check your speciation parameters and the importance factors for speciation. Sometimes there are too many species, sometimes there are too few. The dropoff age...
Is there any study showing a comparison between NEAT and the Cascade Correlation algorithm? Or could Stanley himself points out important differences and...
Having problems with message search?
Fill out this form to ensure your group is one of the first to be migrated to the new message search system.
