<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 3.2//EN"> <HTML> <HEAD> <META HTTP-EQUIV="Content-Type" CONTENT="text/html; charset=iso-8859-1"> <META NAME="Generator" CONTENT="MS Exchange Server version 6.5.7654.12"> <TITLE>RE: [Nnhep] Proposal for the meeting</TITLE> </HEAD> <BODY>  Dear all, i am traveling to CERN today from Montenegro. Unfortunately, i will be busy with the CLIC workshop on both Thursday and Friday the whole day and thus i am not sure i could join you at 29.05.09 11.00-12.00. I will contact you once at CERN. Best, Tomas -----Original Message----- From: nnhep-bounces@projects.hepforge.org on behalf of Lev Dudko Sent: Tue 5/26/2009 12:02 PM To: nnhep@projects.hepforge.org Cc: Konstantin.Toms@cern.ch Subject: [Nnhep] Proposal for the meeting      Dear Colleagues,   we have discussed privately, the possible time of the first meeting dedicated to NNHEP proposal. Looks like 3 people are in CERN now. Sasha can be connected with skype.     The proposed time is 29.05.09 11.00-12.00  If you want to change the time, lets discuss it.     The first version of the questions to discuss is the following: 1. We have to define the structure of interfaces between different modules and understand what should be the input of the module and the output.     Let me remind, what I mean the module, this is one independent component e.g. one neural network or some other algorithm which has some input and some output, internal set of the parameters and some way of graphical representation of what is happened with the input in this module. 2. We can start to discuss what modules (algorithms) we would like to see as the components of the system and what parts we can take from the available external code. The first set could be the following: Feed-Forward Neural Network, Bayesian Neural Network, Boosting algorithm, Decision trees, Self Organizing Maps, Likelihood functions, ...       Some of the modules are realized already in the exist packages (e.g. Feed-Forward NN training with perfect optimizations and training methods are realized in MLPfit package; BNN is realised in RuleFit, etc.) 3. Practical realization of the discussed plans and how we plan to proceed with the code development.         Please, correct/add the set of questions above and provide your opinion.               Best regards,                     Lev --  Lev V. Dudko            e-mail:dudko@fnal.gov  t. +7(495)9395881      <A HREF="http://top.sinp.msu.ru">http://top.sinp.msu.ru</A> </BODY> </HTML>