... [thinking cap on] With a 3x3 grid there are 14 different possible sets of inputs right, so 9 individual squares + 4 (groups of 4) + 1 for the whole 9...
Hi, Intel seem to be offering free copies of some of their Linux development products for non-commercial use. Their definition of "non-commercial" is a bit...
... Here are some results using the very basic test domain John described. This is after about 40k gens, say 30-45 mins (not keeping track of time)... Coords ...
I updated the graphs in my phased search write-up to reflect the changes in SharpNEAT, following the discussions about the slightly odd looking graph with...
... Oh yes, did I not mention that ;-) But.... acherly the number of dimensions is a thing that gets fuzzy as the case gets complex. For example, consider a...
[snip...] ... I think that's a really good insight. However, there's a downside as well. One of the things about higher-dimensional spaces is that our...
[snip...] ... I'm not sure that I agree. Hmm, let me rephrase, I agree that NEAT is an exciting technique and well worthy of all the exploration we can get. ...
... I guess you are talking about the common, short repetitive sequences found in the Junk DNA? They are not used as labels. It's purely position on the...
Hi, This is great! Does the net have any input to tell it how far zoomed it is? How do you decide when the network has reached a decision? Ian Living@Home -...
... Yeah...the network has inputs indicating its x and y position, zoom factor, and rotation. ... We've tried some different approaches to this. At first we...
... Err, you are using neurones which can output more than 1? Or scaling? Also, do you have bias on your nodes, I expect you do, but I thought I'd check...
... Have you tried the method I used in my shape-recogniser, where the network sets a particular output when it's reached a decision? If it doesn't reach a...
... No, we haven't tried that method, but it's something to keep in mind. Our thinking was that an averaged confidence over many steps would yield a more...
... Good catch. The output neuron is a summation node, not a sigmoid. ... Yep! ... It might, but it won't help with where I want to get to. For enlargement,...
This was linked from slashdot today, I haven't read it yet but a quick scan down the article shows pictures of neural nets and other cool diagrams showing how...
... Agreed. So the next step is to evaluate it multiple times per generation with different inputs each time. Assign the 9 pixel inputs a random number from...
... software technique ... hope of ... very like ... the overall ... hugely ... (e.g. each ... overall fitness ... direct ... and all ... benchmark. I ... ...
In the usual GA view of evolution, the issue of genes that are functionally dependent being inherited together seems to be considered significant. However, I...
... handle ... scaling? ... Is there any particular reason you don't use a sigmoid at the outputs. After all, output can always be scaled to any range. But ...
Colin, Here are some comments on your writeup. I think it's great work and very helpful. It seems from your preliminary results that phased search can...
... The raw inputs are scaled before entering them into the network. The output is then scaled back by reversing the scaling using the same parameters. ...
Ian, These emails are getting out of hand so I'm gonna [foolishly] try and merge them... ... Ideally you would do as nature does and increase population so...
... I may be getting confused. I think I read an article about finding genes within all the data generated by the human genome project, I think there are...
... I agree with what you're saying here, but consider this. We have a NEAT genome that describes a network which we can break down into seperate sub-networks,...
At 17:49 02/09/2004, you wrote ... Talking about "genetic" techniques in general, there's two related phenomena: 1) being able to draw equivalences between...
... Yes, I was in the discussion we had here a few months ago. ISRT the Modular NEAT paper did use labels, but they were explicit, not evolving (e.g. the label...
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.
