Sample Efficient Learning with Feature Selection for Factored MDPs

[1703.03454] Sample Efficient Feature Selection for Factored MDPs

In reinforcement learning, the state of the real world is often represented by feature vectors. However, not all of the features may be ...

In reinforcement learning, state is often represented by feature vectors. Prior sample complexity bounds scale with the complexity of all features.

Sample Efficient Feature Selection for Factored MDPs - arXiv

Much preferable are reinforcement learning algorithms whose sample complexity scales only with the number of relevant features needed to learn ...

[PDF] Sample Efficient Feature Selection for Factored MDPs

Found. Trends Mach. Learn. 2018. TLDR. This manuscript provides an introduction to deep reinforcement learning models ...

Sample Efficient Feature Selection for Factored MDPs - ResearchGate

We propose Feature Selection Explore and Exploit (FS-EE), an algorithm that automatically selects the necessary features while learning a Factored Markov ...

[PDF] Sample Efficient Learning with Feature Selection for Factored ...

2 Citations · Towards Minimax Optimal Reinforcement Learning in Factored Markov Decision Processes · Improved Exploration in Factored Average-Reward MDPs.

Sample Efficient Feature Selection for Factored MDPs - DeepAI

03/09/17 - In reinforcement learning, the state of the real world is often represented by feature vectors. However, not all of the features ...

Efficient Structure Learning in Factored-state MDPs - Carlos Diuk

A factored-state MDP is one whose states are repre- sented as a vector of distinct components or features. Dy- namic Bayesian networks (DBNs) and decision trees ...

Efficient Solution Algorithms for Factored MDPs

Clearly, feature selection is an important issue for essentially all areas of learning and approximation. ... Basis Function Selection: In this simple example ...

Efficient Solution Algorithms for Factored MDPs - Stanford AI Lab

feature selection is an important issue for essentially all areas of learning and approximation. We offer some simple methods for selecting good features ...

Efficient Structure Learning in Factored-State MDPs

A factored-state MDP is one whose states are repre- sented as a vector of distinct components or features. Dy- namic Bayesian networks (DBNs) and decision trees ...

Efficient Reinforcement Learning in Factored MDPs - UPenn CIS

variable's behavior is identical for several actions. 3 Overview of the Original E. 3. Since our algorithm for learning in DBN-MDPs will be ...

EFFICIENT REINFORCEMENT LEARNING IN FACTORED MDPS ...

Q-learning with UCB exploration is sample efficient for infinite-horizon MDP. arXiv preprint arXiv:1901.09311, 2019. Yonathan Efroni, Shie Mannor, and ...

Sample-Efficient Reinforcement Learning for Linearly ...

This matches the minimax optimal lower bound (up to a logarithm factor) established in [YW19, Theorem 1] for feature-based MDP. In comparison, for tabular MDP ...

Sample-Efficient Reinforcement Learning in the Presence of ...

Zhaohan Daniel Guo and Emma Brunskill. Sample efficient feature selection for factored MDPs. arXiv preprint arXiv:1703.03454, 2017. 15 ...

Reinforcement Learning in Factored MDPs: Oracle-Efficient ...

We provide two oracle-efficient algorithms, DORL (Discrete Optimism RL) and PSRL. (Posterior Sampling RL), with near-optimal frequentist regret bound and ...

Efficient solution algorithms for factored MDPs - ACM Digital Library

A central element of our algorithms is a novel linear program decomposition technique, analogous to variable elimination in Bayesian networks, which reduces an ...

Sample-Efficient Reinforcement Learning for Linearly ... - OpenReview

and γ ∈ (0, 1) is the discount factor of the MDP. The results is applicable to the tabular MDPs by taking the coordinate basis with K = |S| × |A|. Both sample.

(PDF) Automatic Feature Selection for Model-Based Reinforcement ...

performance and reduce the computational expense of planning. Keywords-Reinforcement learning; feature selection; factored. MDPs.

Automatic Feature Selection for Model-Based Reinforcement ...

REINFORCEMENT LEARNING IN FACTORED MDPS. A Markov decision process (MDP) can ... Koller, “Efficient reinforcement learning in factored. MDPs,” in ...