Mash-WP6 - A stochastic dynamic programming framework
01 January 2006

Person in charge : TEYTAUD Olivier


(95000 lines of code, C++) OpenDP is a stochastic dynamic programming framework, including optimisation, sampling, learning and dynamic programming algorithms.
OpenDP is under GPL license. While the number of users is still restricted, some modules have been reused in widely distributed programs (in particular some quasi-random sequences in the Gnu Scientific Library (see below) and the quasi-random Evolution Strategies in the very widely distributed Beagle). The OpenDP platform specifically includes:

• random and quasi-random number generators;


• support vector machines with several kernels, neural networks, decision trees, dots (including Weka);


• cross-validation and heuristics for the choice of parameters;


• various non-linear optimisation tools (evolution strategies, genetic algorithms, direct search methods, gradient-based techniques).

More information: http://opendp.sourceforge.net/wiki/

Software - Licence : GPL



Research activities
  Machine learning
  Optimization

Members
  TEYTAUD Olivier
  MARY Jérémie
  GELLY Sylvain

Group
  Learning and Optimization

BSP++
The C++ Bulk Synchronous Parallelism Library

TAXOMAP ALIGNMENT
A prototype to automate semantic mappings between taxonomies

FR1155729
Procédé pour l’extinction de routeurs dans un réseau de communications et routeur mettant en œuvre ce procédé