Yawn (Yet Another {Weird | Wacky | Witty | Wonderful | W*} Network) is a Java™ framework for training and testing neural networks independently from the neural model and the test environment being used.

The models implemented so far are AppART , GasART , fuzzy ARTMAP and Multi-Layer Perceptron (MLP). Note that the MLP implementation, although fully workable, is a reference implementation meant for educational use and might yield less-than-optimal results.

Yawn does not perform any kind of statistical pre- or post-processing neither any results analysis. These tasks are left over to the chosen test environment. Currently, Yawn is capable of interacting with the DELVE test environment and with plain text files. A Matlab™ and a xml-file test environment are currently under development.

URL: http://yawn2.sourceforge.net/
Licence: GPL