I know there are a bunch of TOTAs in engineering, software, etc., so might as well take advantage of the manfest here...

Any of you know a thing or two about predictive algorithms? In particular I'm looking for info on neural networks and Fourier series (curve fitting, NOT FFTs), but other ideas are welcome too.

I'm not looking for newbie tutorials; I understand all the concepts behind FS and NNs, I just have no clue how to even get started on implementing a real one. If anyone has even the tiniest bit of experience with these things, or knows someone who does, or has a book that they've never read on it, please let me know - code language doesn't matter, I understand pretty much all of them.



...and I know a bunch of random people who are not geeks are going to read this because they think it's another rant. Well it isn't dagnabbit, sometimes I've got better things to do!

___________________
My party schedule:
2009-02-21 - DJ Attention @ I'm So Popular
2009-06-18 - DJ Annoying @ People Need To Know Where I'll Be
2012-11-32 - DJ Insufferable ɸ Or At Least the Stalkers I Complain About
2048-06-66 - Spastic & Whocares ¶ Although I'm Actually Flattered
9999-45-81 - Tweaker Gimp ☼ I Probably Won't Even Go To This But I Have To Make Sure I Fill Up All The Available Space Here

quote:

Originally posted by DigiNut Fourier series

Fourier and La Grange are my whores!

Dig these doods for info (contact them and ask as much as you want). Yay IEEE!

http://www.ieee-nns.org/

___________________
- rabbit.joker [funny¿rabbit] | www.rabbitjoker.com |www.ddtt.org

Dark Dirty Tech Tribal. | Hands in air (trance) and feet on the floor (house).

Hrm, well if I was actually in class rather than twiddling my thumbs, I'd be taking an entire course on this right now.

I'd give you my differential equations book for a bit, but to be honest, it's pretty much useless, even to me. Besides, it only has one chapter on fourier series, and won't cover anything in great detail.

I'd recommend getting down to U of T's science and engineering library [or whatever library they have that has all the maht and science stuff] and asking one of the librarians. I know for a fact Mac has some wicked alternate sources for learning math, it's jsut a matter of getting your ass out to the library and asking.

___________________
Not Everyone Understands House Music, It's a Spiritual Thing, a Body Thing, a Soul Thing

quote:

Originally posted by DigiNut ...and I know a bunch of random people who are not geeks are going to read this because they think it's another rant. Well it isn't dagnabbit, sometimes I've got better things to do!

lol.. it's all bout tha zener diodes!

___________________
Palm Trees > Pine Trees , Sand > Snow

quote:

Originally posted by rabbitjoker Fourier and La Grange are my whores! Dig these doods for info (contact them and ask as much as you want). Yay IEEE! http://www.ieee-nns.org/

Hey RJ - thanks for the link. Have you spoken to any of those dudes in particular that you know would be helpful? It's probably not practical for me to actually take out a membership - I can always try all of them and see who doesn't mind helping out a total stranger but I might as well check with you first to see if I can narrow down the list.

Also to the other guys, keep in mind that I did complete four years of EE so I know all the textbook theory behind differential equations and Fourier series. This is more about how to actually put that theory into practice as a curve-fit/predictive algorithm in code - sadly, university programs and books are sorely lacking in such practical knowledge.

I bet somebody on here has done it before. Durafei, I'm looking at you.

Thanks again guys

___________________
My party schedule:
2009-02-21 - DJ Attention @ I'm So Popular
2009-06-18 - DJ Annoying @ People Need To Know Where I'll Be
2012-11-32 - DJ Insufferable ɸ Or At Least the Stalkers I Complain About
2048-06-66 - Spastic & Whocares ¶ Although I'm Actually Flattered
9999-45-81 - Tweaker Gimp ☼ I Probably Won't Even Go To This But I Have To Make Sure I Fill Up All The Available Space Here

I haven't spoken to anybody at the IEEE in probably 4 years...

However, I would suggest talking to this guy - his company specializes in exactly what you are looking for:

David B. Fogel
Natural Selection, Inc.
3333 N. Torrey Pines Ct., Suite 200
La Jolla, CA 92037, USA
phone: (+1 858) 455 6449
fax: (+1 858) 455 1560
email: dfogel .a_t. natural-selection.com
www: www.natural-selection.com

___________________
- rabbit.joker [funny¿rabbit] | www.rabbitjoker.com |www.ddtt.org

Dark Dirty Tech Tribal. | Hands in air (trance) and feet on the floor (house).

quote:

Originally posted by DigiNut I know there are a bunch of TOTAs in engineering, software, etc., so might as well take advantage of the manfest here... Any of you know a thing or two about predictive algorithms? In particular I'm looking for info on neural networks and Fourier series (curve fitting, NOT FFTs), but other ideas are welcome too. I'm not looking for newbie tutorials; I understand all the concepts behind FS and NNs, I just have no clue how to even get started on implementing a real one. If anyone has even the tiniest bit of experience with these things, or knows someone who does, or has a book that they've never read on it, please let me know - code language doesn't matter, I understand pretty much all of them. ...and I know a bunch of random people who are not geeks are going to read this because they think it's another rant. Well it isn't dagnabbit, sometimes I've got better things to do!

If you want to write a neural net then your going to have to give a little more context on what your trying to do. With what data do you plan to train it? Good and accurate neural nets are typically run against thousands of training data.

-Your Mother

quote:

Originally posted by DigiNut Hey RJ - thanks for the link. Have you spoken to any of those dudes in particular that you know would be helpful? It's probably not practical for me to actually take out a membership - I can always try all of them and see who doesn't mind helping out a total stranger but I might as well check with you first to see if I can narrow down the list. Also to the other guys, keep in mind that I did complete four years of EE so I know all the textbook theory behind differential equations and Fourier series. This is more about how to actually put that theory into practice as a curve-fit/predictive algorithm in code - sadly, university programs and books are sorely lacking in such practical knowledge. I bet somebody on here has done it before. Durafei, I'm looking at you. Thanks again guys

If your doing curve fitting, it sounds like you want to build a fuzzy system instead.

-Your Mother

My CPU is a neural-net processor, a learning computer!

quote:

Originally posted by goodnet My CPU is a neural-net processor, a learning computer!

quote:

Originally posted by goodnet My CPU is a neural-net processor, a learning computer!

lol..

___________________
Palm Trees > Pine Trees , Sand > Snow

quote:

Originally posted by Your Mother If you want to write a neural net then your going to have to give a little more context on what your trying to do. With what data do you plan to train it? Good and accurate neural nets are typically run against thousands of training data. -Your Mother

It's for forecasting utility consumption. We have literally millions of data points to train it on, so I don't think that's a big issue. I already have a plan for how to train it, I just need to know how to code it.

quote:

If your doing curve fitting, it sounds like you want to build a fuzzy system instead.

The curve fitting only really applies to the Fourier series. I want to apply it as a different type of forecasting, because the data is *somewhat* cyclical and I tried a few curve-fitting programs that came up with good matches using the FS. I have no idea how those programs generate the functions though. The idea is to have a few different algorithms running - we already have 2 of them, but we need just *that* much more precision.

It's definitely a fuzzy system - the algorithms I used originally were based heavily on statistics and probabilities.

___________________
My party schedule:
2009-02-21 - DJ Attention @ I'm So Popular
2009-06-18 - DJ Annoying @ People Need To Know Where I'll Be
2012-11-32 - DJ Insufferable ɸ Or At Least the Stalkers I Complain About
2048-06-66 - Spastic & Whocares ¶ Although I'm Actually Flattered
9999-45-81 - Tweaker Gimp ☼ I Probably Won't Even Go To This But I Have To Make Sure I Fill Up All The Available Space Here