Direct policy gradient methods for reinforcement learning are a successful approach for a
variety of reasons: they are model free, they directly optimize the performance metric of …

We consider approximate dynamic programming in $\gamma $-discounted Markov decision
processes and apply it to approximate planning with linear value-function approximation …

Reinforcement learning (RL) is a classical tool to solve network control or policy optimization
problems in unknown environments. The original-learning suffers from performance and …

Reinforcement learning is a framework for interactive decision-making with incentives
sequentially revealed across time without a system dynamics model. Due to its scaling to …

We study the problem of optimizing a recommender system for outcomes that occur over
several weeks or months. We begin by drawing on reinforcement learning to formulate a …

This paper proposes a framework for learning the most concise MDP model of a continuous
state space dynamic system from observed transition data. This setting is encountered in …

We apply functional acceleration to the Policy Mirror Descent (PMD) general family of
algorithms, which cover a wide range of novel and fundamental methods in Reinforcement …

State abstraction has been an essential tool for dramatically improving the sample efficiency
of reinforcement-learning algorithms. Indeed, by exposing and accentuating various types of …

State Abstraction is a methodology that aims to simplify planning problems and enable
planners to deal with more complex environments. It is a useful tool that helps Artificial …

In this thesis we present contributions to the theoretical foundations of large-scale
reinforcement learning (RL) with linear function approximation, with a focus on establishing …