View Document Preview and Link Document Date: 2014-02-17 17:15:07 Open Document File Size: 2,95 MB Share Result on Facebook
City Probabilistic Model / C. Gradient / / Company Autonomous Systems / MIT Press / PrimeSense / / Country Germany / Jordan / United Kingdom / / Currency USD / / Facility University of Stuttgart / University of Washington / Imperial College / Max Planck Institute / / IndustryTerm grid search / policy-search approach / classical policy search / multi-task policy search / tasks using policy search / policy search framework / policy search / policy search methods / / Organization University of Washington / Max Planck Institute for Intelligent Systems / University of Stuttgart / Imperial College London / Department of Computer Science / MIT / Department of Computing / Department of Computer Science and Engineering / / Person Marc Peter / / Position teacher / NN-IC controller / nonlinear controller / learned controller / resulting controller / hierarchical RW-IC controller / single controller / learned multi-task controller / MTPS0 controller / single controller for multiple tasks jointly / controller / MTPS+ controller / / ProgrammingLanguage J / / Technology resulting algorithm / GPS / / URL http / SocialTag 
Multi-Task Policy Search for Robotics Marc Peter Deisenroth1,2 , Peter Englert3 , Jan Peters2,4 , and Dieter Fox5 Abstract— Learning policies that generalize across multiple tasks is an important and challenging resear