2010/05/11 by Marek Petrik, Gavin Taylor, Petrik, Marek +6 · 2 citations
Computer Science · Engineering · #Adaptive Dynamic Programming Control #Artificial Intelligence (cs.AI) #Control Systems and Identification #FOS: Computer and information sciences #Reinforcement Learning in Robotics #Water resources management and optimization
paper · pdf · doi:10.48550/arxiv.1005.1860
openalex publication_date 2010/05/11 · openalex created_date 2019/06/27 · openalex updated_date 2026/07/28
Approximate dynamic programming has been used successfully in a large variety\nof domains, but it relies on a small set of provided approximation features to\ncalculate solutions reliably. Large and rich sets of features can cause\nexisting algorithms to overfit because of a limited number of samples. We\naddress this shortcoming using L1 regularization in approximate linear\nprogramming. Because the proposed method can automatically select the\nappropriate richness of features, its performance does not degrade with an\nincreasing number of features. These results rely on new and stronger sampling\nbounds for regularized approximate linear programs. We also propose a\ncomputationally efficient homotopy method. The empirical evaluation of the\napproach shows that the proposed method performs well on simple MDPs and\nstandard benchmark problems.\n