# Reference request for neural network programming in Mathematica

I'm looking for a good reference/book on programming neural networks in mathematica. I've been working through Freeman's "Simulating Neural Networks with Mathematica," but it is from 1994 so is quite dated. Is anyone familiar with a more recent book on the subject?

For background, I'm very comfortable with pure mathematics, somewhat comfortable with general programming, and have been teaching myself to be proficient in mathematica. I'd like to play around with implementing neural networks on mathematica for use in forecasting of high frequency economic data, so anything with a finance bent is an especially welcome reference.

-
Have you checked out the Neural Networks add-on package? Its pretty expensive but there might be relevant documentation or authors who could get you started. wolfram.com/products/applications/neuralnetworks –  Andy Ross Feb 26 '12 at 2:57
Thanks, I had seen that, but figured the documentation would be focused on specifically using the neural-nets package. It is a good idea to go through some of the authors and see if I can find anything they've written not specifically tailored to the package. I'm just not up for spending \$900 for the add-on at this point since this is purely for my own entertainment/education. –  user600 Feb 26 '12 at 3:25
This blog deals with datamining in Mathematica. There are some nice posts about neural nets as well. –  Sjoerd C. de Vries Feb 26 '12 at 21:54
Wolfram occasionally runs courses in this kind of thing: See: wolfram.com/services/education/courses/m330.html –  Jagra Feb 28 '12 at 1:58
Is the NeuralNetworks package still under development? The example here (library.wolfram.com/examples/exchangerate) and the official documentation (media.wolfram.com/documents/NeuralNetworksDocumentation.pdf) seem to use <<LinearAlgebraMatrixManipulation which is obsolete! –  my account_ram Oct 27 '13 at 17:01

I've had an interest (as one can see in my other posts) in a wide range of distributed processing and parallel computing approaches and while not seen in any of my posts machine learning approaches as well. I looked at neural networks some years ago, and while they didn't suit the problems I worked on at the time I remembered the article Duncan and Tweney wrote as useful. A couple of others might also prove useful.

Three references follow:

AI AND STATISTICAL APPLICATIONS Mathematica: A flexible design environment for neural networks

From the Journal: Behavior Research Methods, Instruments, & Computers 1997,29 (2). 194-199

From 1997, a few years more recent than Freeman's.

Abstract:

Several neural networks were developed in Mathematica in order to explore the role of "spiky" neurons in neural network memory simulations. Using Mathematica for this task confirmed its value as a powerful tool for neural network development: It exhibited distinct advantages over other environments in programming ease, flexibility of data structures, and the graphical assessment of network performance.

One of its authors: Sean C. Duncan has moved from Bowling Green University to Miami University. He has a website: http://se4n.org/

Its other author: Ryan Tweney remains at Bowling Green University and has his own website: http://personal.bgsu.edu/~tweney/

You can find contact information for each of them on their respective websites. I've always found academics generous with what they know. The article or contacting them might lead you to better sources of information on this.

Mathematica Neural Networks package.

You can download the pdf of the manual for the Mathematica Neural Networks package. Pretty extensive, indeed.

The Power of Neural Networks

A review in which Brian Cogan briefly assesses NeuroSolutions from NeuroDimension, and Neural Networks, a Mathematica add-on, from Scientific Computing World March/April 2003

http://www.neurosolutions.com/resources/scw.pdf

-