MLPfit:a tool to design and use Multi-Layer Perceptrons
(Artificial) Neural Networks
Neural Networks in HEP
The most wi(l)dely used: the Multi-Layer Perceptron
2 theorems
About learning ...
Remark: derivative of sigmoid function
Learning methods
The traditional learning method: stochastic minimization
Stochastic minimization (continued)
Other methods
The Hybrid method
Comparison on a toy problem
Comparison on a real problem
The MLPfit package
Software process
Performances
Using MLPfit as a standalone package
What is MLP_lapack.c ?
Calling MLPfit from your own code
Calling MLPfit from LabVIEW
Using MLPfit in PAW
MLPfit in PAW: modified vec/fit command
MLPfit in PAW: interface to Ntuples
Documentation ...
Conclusions
We are willing to ...
Email: jerome@hep.saclay.cea.fr
Download presentation source