Begin3
Title:          mozg
Version:        0.11.3
Entered-date:   19OCT99
Description:    fast C++ neural network library. Supports training, running,
                saving and loading of networks with an arbitrary number of
                layers and nodes.
                Learning rules: Backprop with extensions (chunk or on-line),
                Quickprop, Rprop.
                Neuron output functions: sigmoid, th, atg, linear, RBF or
                desirable one may be inserted easy. Any combination of them
                for different layers is possible.
                Energy function: summed squared error, cross-entropy, 
                log-squared error.
                Kernel of the simulator is a library of optimised vector &
                matrix operations, so it may be replaced by faster one (may be
                hardware implemented) to sped up the simulator. Optimisation
                technique has been described.
                Library is used in DELPHI(CERN) experiment data analysis.
Keywords:       neural network, Backpropagation, Quickprop, Rprop, library, C++
Author:         filin@desert.ihep.su (Alexy Filin)
Maintained-by:  filin@desert.ihep.su (Alexy Filin)
Primary-site:   sunsite.unc.edu /pub/Linux/science/ai
                1.5k  lsm-template
                158k mozg-0.11.3.tar.gz
Alternate-site: Linux.ihep.su /pub/Linux/local/mozg-0.11.3.tar.gz
Original-site:  Linux.ihep.su /pub/Linux/local/mozg-0.11.3.tar.gz
Platforms:      UNIX-like
Copying-policy: LGPL
End