Albrecht Schmidt Manchester Metropolitan University Department of Computing September 1996 The neural network Software *************************** This package was developed as a part of my MSc Project at the Manchester Metropolitan University 1996. The Program was used to compare different network structures and types. The source code can be used as a construction kit for artificial neural networks. The software was created as a prototype to evaluate different theorectical ideas, so don't tell that it is slow, or that the the user-interface is not graphical! The package contains three types of neural networks: a standart Backpropagation implementation, a logical neural network, and a modular neural network. The implementation is objected-oriented. To install the software: (you have already used tar -xvf and uncompress) change to the directory impl: cd impl complile the software: make change back one level: cd .. add the directory bin to the path: set path=($path /bin) to find out what is: pwd The sub-directories: ./demo A demo how to use the software ./bin The executable programs are in this sub-directory ./impl The source code is in this sub-directory and in its subdirectories ./impl/tools examples of converter for the data files ./testdata Data to use the network is in this sub-directory To use the programs: start the program: bpnet for the Backpropagation network logicnet for the logical neural network modnet for the modular neural network Use the option ew to create a new network, then train the network using the training data set, after this store the network. To test the performance with the menu topic

erformance. The information about training time, recognition perfomance, etc. is stored in the log-file, the propagated error is stored in the error file. ---------------------------------------------------------------------------- If you have any questions please contact me: smail: Albrecht Schmidt Beuerlbacher Hauptstrasse 56 74564 Crailsheim - Beuerlbach Germany email: aschmidt@hydra.infromatik.uni-ulm.de