Usable Neural Network Libraries?

hugin asks: "I was just wondering if anyone had the skinny on open-sourced neural network libraries. I am looking for one for an academic use as a component of a flexible software based DSP application that I am putting together under Windows. Something with separate callable instantiation, training and running (for a frozen network) functions would be great. Something with the functionality to save a network to a file for later opening would be even better." bartok is looking for something similar, but he also wants "to deploy neural network infrastrucures on Linux or any open platform (i.e., Web Apps)."

Re:Try SNNS

[mlc]

For the truly lazy, a clickable link [uni-stuttgart.de] to the site mentioned above.

NNO

[jefferson]

Take a look at Neural network objects [ruhr-uni-bochum.de] by Johannes Steffens. It's a C++ class library that supports both supervised and unsupervised learning networks including ordinary backpropagation (or MLP, multi-layer perceptron) nets, as well as Kohonen feature maps (KFM, aka SOM) and learning vector quantization.

It also supports several growing network architectures developed by Bernd Fritzke [tu-dresden.de] et al. There's also a Java demo [ruhr-uni-bochum.de] of growing networks, with code available. It's fun to change the input probability distribution and watch the nets adapt on the fly!

J.