Pavel Surynek's Academic Page | NN-Lab


Author:Pavel Surynek

NN-Lab is a program for creating, training and using multi-layer nerural networks. It was written in C. The program communicates with user through its own command line, where the user can enter commands provided by the program. Although a help for every command is available, it is assumed that the user has some basic knowledge about the neural networks. The most important feature of NN-Lab are algorithms for training the nets. There are implemented two algorithms: a classical backpropagation algorithm and a genetic algorithm, both are available in a few modifications.

Following pictures explain the usefullness of neural networks. Briefly said, the multi-layer neural netwoks worth consideration for apllications where some level of robustness of a computation process is necessary, in other words, where the computation process should be imune to some noise in the input data.

The input picture is perferct, the recognition via neural network works well:

The input picture contains some noise, but the recognition via neural network works well again:

Here you can download NN-Lab:

  • NN-Lab 1.02  Linux/Unix version
  • NN-Lab 1.02  Win32/MS-Visual Studio version

  • Keywords: neural networks, backpropagation, genetic algorithms