Grid'5000 user report for

Jump to: navigation, search

User information

( user)
More user information in the user management interface.


  • Cross-entropy method for Reinforcement Learning (Programming) [in progress]
    Description: We are investigating how cross-entropy (CE) method can be used in reinforcement learning for finding the optimal value function and/or policy. CE is a Monte Carlo approach to combinatorial and continuous multi-extremal optimization and importance sampling. The experiments involve iteratively creating a set of candidate samples (approximation of the value function or a policy) according to a parameterized (random) distribution, and then, based on the performances of the samples the parameters are updated using the cross-entropy principle with the aim of producing a better sample in the next iteration. Grid5000 is used to parallelize the process of evaluating the performance of candidate samples.
    Results: In progress.



    Success stories and benefits from Grid'5000

    last update: 2008-02-18 16:10:21