neural networks for finding patterns in sequences of numbers

Dear all, I've been reading that there is a "Neural Network Toolbox" that can help in finding patterns, and clustering data, so I'm wondering whether it can help me with my job or not, and how difficult it will be to make neural networks work for me.

So my problem is that I have a [187, 301] matrix of sequences of numbers in which every row represents a kind of biochemical signal. There are not described patterns for this type of measurement so I need to start from zero.

Now, I don't know how long are the "patterns" I'm looking for, neither the shape of them. Every number is related to a position (like in the DNA) not to a time series.

What I have been doing is using xcorr at zero lags to find the most similar "patterns" for different lengths looping through all the sequences. This way I find common "patterns" but is SO inefficient and time consuming (but precise I have to say) that I'm trying to find a different approach.

As far as I read neural networks can be trained for doing the type of thing I'm looking for. I don't have the neural networks toolbox, but a trial is available for download so I'll try it if somebody confirms it is going to be useful for my specific needs.

Thanks in advance,

Diego

[Diego]

Any advice?

Answer 1

[Walter Roberson]

Sounds like some of the research we did about 13 to 15 years ago; we were working with functional MRI (fMRI) and had to determine responses without knowing what (or when) the inputs were. The algorithms did not depend upon it being MRI -- or upon it being a time-course; it was also able to recognize lags between locations.

Let's see... the paper is <http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.33.7510&rep=rep1&type=pdf> and it appears that there is still a copy of the users' manual around, http://homepage.usask.ca/~ges125/fMRI/Evident_manual.pdf for whatever good that will do. Unfortunately I believe the source itself is now lost (or at the very least, difficult and expensive to retrieve if we still have the media around at all.)

[Diego]

Nice!!!

I'll read them through.

Best,

Diego