We study the sample complexity of model-based reinforcement learning in general contextual decision processes. We design new algorithms for RL with an abstract model class and analyze their statistical properties. Our algorithms have sample complexity governed by a new structural parameter called the witness rank, which we show to be small in several settings of interest, including Factored MDPs and reactive POMDPs. We also show that the witness rank of a problem is never larger than the recently proposed Bellman rank parameter governing the sample complexity of the model-free algorithm OLIVE (Jiang et al., 2017), the only other provably sample efficient algorithm at this level of generality. Focusing on the special case of Factored MDPs, we prove an exponential lower bound for all model-free approaches, including OLIVE, which when combined with our algorithmic results demonstrates exponential separation between model-based and model-free RL in some rich-observation settings.

We study the computational tractability of provably sample-efficient (PAC) reinforcement learning in episodic environments with high-dimensional observations. We present new sample efficient algorithms for environments with deterministic hidden state dynamics but stochastic rich observations. These methods represent computationally efficient alternatives to prior algorithms that rely on enumerating exponentially many functions. We show that the only known statistically efficient algorithm for the more general stochastic transition setting requires NP-hard computation which cannot be implemented via standard optimization primitives. We also present several examples that illustrate fundamental challenges of tractable PAC reinforcement learning in such general settings.

Most contextual bandit algorithms minimize regret against the best fixed policy, a questionable benchmark for non-stationary environments that are ubiquitous in applications. In this work, we develop several efficient contextual bandit algorithms for non-stationary environments by equipping existing methods for i.i.d. problems with sophisticated statistical tests so as to dynamically adapt to a change in distribution. We analyze various standard notions of regret suited to non-stationary environments for these algorithms, including interval regret, switching regret, and dynamic regret. When competing with the best policy at each time, one of our algorithms achieves regret $\mathcal{O}(\sqrt{ST})$ if there are $T$ rounds with $S$ stationary periods, or more generally $\mathcal{O}(\Delta^{1/3}T^{2/3})$ where $\Delta$ is some non-stationarity measure. These results almost match the optimal guarantees achieved by an inefficient baseline that is a variant of the classic Exp4 algorithm. The dynamic regret result is also the first one for efficient and fully adversarial contextual bandit. Furthermore, while the results above require tuning a parameter based on the unknown quantity $S$ or $\Delta$, we also develop a parameter free algorithm achieving regret $\min\{S^{1/4}T^{3/4}, \Delta^{1/5}T^{4/5}\}$. This improves and generalizes the best existing result $\Delta^{0.18}T^{0.82}$ by Karnin and Anava (2016) which only holds for the two-armed bandit problem.

We study and empirically optimize contextual bandit learning, exploration, and problem encodings across 500+ datasets, creating a reference for practitioners and discovering or reinforcing a number of natural open problems for researchers. Across these experiments we show that minimizing the amount of exploration is a key design goal for practical performance. Remarkably, many problems can be solved purely via the implicit exploration imposed by the diversity of contexts. For practitioners, we introduce a number of practical improvements to common exploration algorithms including Bootstrap Thompson sampling, Online Cover, and $\epsilon$-greedy. We also detail a new form of reduction to regression for learning from exploration data. Overall, this is a thorough study and review of contextual bandit methodology.

This paper studies the evaluation of policies that recommend an ordered set of items (e.g., a ranking) based on some context---a common scenario in web search, ads, and recommendation. We build on techniques from combinatorial bandits to introduce a new practical estimator that uses logged data to estimate a policy's performance. A thorough empirical evaluation on real-world data reveals that our estimator is accurate in a variety of settings, including as a subroutine in a learning-to-rank task, where it achieves competitive performance. We derive conditions under which our estimator is unbiased---these conditions are weaker than prior heuristics for slate evaluation---and experimentally demonstrate a smaller bias than parametric approaches, even when these conditions are violated. Finally, our theory and experiments also show exponential savings in the amount of required data compared with general unbiased estimators.

This paper studies the evaluation of policies that recommend an ordered set of items (e.g., a ranking) based on some context---a common scenario in web search, ads, and recommendation. We build on techniques from combinatorial bandits to introduce a new practical estimator that uses logged data to estimate a policy's performance. A thorough empirical evaluation on real-world data reveals that our estimator is accurate in a variety of settings, including as a subroutine in a learning-to-rank task, where it achieves competitive performance. We derive conditions under which our estimator is unbiased---these conditions are weaker than prior heuristics for slate evaluation---and experimentally demonstrate a smaller bias than parametric approaches, even when these conditions are violated. Finally, our theory and experiments also show exponential savings in the amount of required data compared with general unbiased estimators.

We investigate active learning with access to two distinct oracles: LABEL (which is standard) and SEARCH (which is not). The SEARCH oracle models the situation where a human searches a database to seed or counterexample an existing solution.

We propose Sketched Online Newton (SON), an online second order learning algorithm that enjoys substantially improved regret guarantees for ill-conditioned data. SON is an enhanced version of the Online Newton Step, which, via sketching techniques enjoys a running time linear in the dimension and sketch size. We further develop sparse forms of the sketching methods (such as Oja's rule), making the computation linear in the sparsity of features. Together, the algorithm eliminates all computational obstacles in previous second order online learning approaches.

We propose and study a new model for reinforcement learning with rich observations, generalizing contextual bandits to sequential decision making. These models require an agent to take actions based on observations (features) with the goal of achieving long-term performance competitive with a large set of policies. To avoid barriers to sample-efficient learning associated with large observation spaces and general POMDPs, we focus on problems that can be summarized by a small number of hidden states and have long-term rewards that are predictable by a reactive function class. In this setting, we design and analyze a new reinforcement learning algorithm, Least Squares Value Elimination by Exploration. We prove that the algorithm learns near optimal behavior after a number of episodes that is polynomial in all relevant parameters, logarithmic in the number of policies, and independent of the size of the observation space. Our result provides theoretical justification for reinforcement learning with function approximation.

Many machine learning applications involve jointly predicting multiple mutually dependent output variables. Learning to search is a family of methods where the complex decision problem is cast into a sequence of decisions via a search space. Although these methods have shown promise both in theory and in practice, implementing them has been burdensomely awkward. In this paper, we show the search space can be defined by an arbitrary imperative program, turning learning to search into a credit assignment compiler. Altogether with the algorithmic improvements for the compiler, we radically reduce the complexity of programming and the running time. We demonstrate the feasibility of our approach on multiple joint prediction tasks. In all cases, we obtain accuracies as high as alternative approaches, at drastically reduced execution and programming time.