I know some people are interested in the indirect encoding-related paper, "Patterns Without Development," that I recently mentioned to the group. However,...
Hi Achim, Unfortunately I do not believe there is a lot of data regarding NEAT's performance on noisy time series. There is reason to believe (based on...
It is true that NEAT is generally not used with separate validation and test sets, but that is only because NEAT is not normally used in supervised domains (it...
There has been significant work showing that a combination of NEAT to optimize topology and initial weights, and an addition learning algorithm to further...
As part of an undergraduate research project, I did implement something on those lines. I worked on a NEAT based approach to predicting a noisy financial time...
Hello all, I saw some messages about financial prediction methods using NEAT and I simply wonder if someone get interesting results using these techniques. In...
Sam, check out the file "ProjectReport.pdf" by Sidhant Dash in the Files section of this group. Sidhant did essentially what you are suggesting (with some...
Sidhant, I have to say that your paper is a great accomplishment for a 3-month undergraduate research project. The scope of this project and amount of...
Hi Sidhant Great paper. I have been trying to do much the same thing, except rather than value prediction, I have been creating a trader that essentially...
All technical indicators are based on some combination of open, close, high, low and volume. I like the approach taken in the book 'Blondie24' - give the...
I hadn't thought about using a value of N as determined by the network - well no in this experiment. I have done it in my chess experiment. Maybe I will do...
Hi Emyr, All Just a quick note to say I've been working on evolving ANNs that give buy,sell,close signals against London shares. The exact set of inputs has...
By 'normalised' what I mean is this.... Get the last N days. Get the highest high and the lowest low over the period. Rescale all the data so that the lowest...
Hello all, I've made some tests some time ago to try to predict the stock market too, but I've made some changes to deal with the input data. All the input are...
Hi, I could do only 20 generations or less for most of my experiments. I understand that with more generations, I could have achieved much better networks. The...
... While I believe it's absolutely essential to transform input data, I'm not a huge fan of this type of normalisation, which tends to compress the bulk of...
Hi All, my name is Aleks. I'm a student at University of Virginia (I work with metal alloys www.virginia.edu/ms). AI is a brand new field to me and I'm...
Hi Aleks, Welcome to the group! That looks an awful lot like ANJI output, so I'm going to assume that's the implementation you're using. I'm one of the ...
Thanks! That was a very helpful and thorough response. It appears that I'm screwing something up in matlab so I'll be looking over that code to find my...
... Hello Peter, My name is Krasimir Trifonov and I am also new in the group. I implemented cross-validation using Ken's and yours suggestion, but there are...
Hello All, I will introduce myself ones again – my name is Krasimir Trifonov. The problem that I would like to try using NEAT is related to the world of...
I've been working on my own implimentation of NEAT (having to fit to a predescribed framework) and while it's mostly working, I'm looking at a couple of...
It sounds like the main issue before you can really start looking at delayed-XOR is to get XOR working more closely to my own NEAT (and others developed...
... Hi Mike, Emyr, All This definitely seems to be the most sound approach to take to me. It does introduce the problem of having to calculate (sample) the...
I haven't had a chance to go back and try the weight adjustment test to start, but I did try the C# NEAT package that already exists to try out the delayed XOR...
Reading through the folding@home web site I see they have been working on a client for ATI graphics cards that achieves "a 20x to 40x speed increase over...
Hi, I just thought I'd pop up and reasure the group that I do still exist... I've been distracted by: 1) summer 2) getting promoted at work 3) a long and...
So I've had a chance to do what Ken suggested below - start from the basic (2+1bias)-1-1 network for the static XOR problem, randomizing the test data, 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.
