More Like this

1. Control Regularization for Reduced Variance Reinforcement Learning

   Cheng, Richard Cheng; Verma, Abhinav; Orosz, Gabor; Chaudhuri, Swarat; Yue, Yisong; Burdick, Joel (January 2019, Proceedings of Machine Learning Research)

   Dealing with high variance is a significant challenge in model-free reinforcement learning (RL). Existing methods are unreliable, exhibiting high variance in performance from run to run using different initializations/seeds. Focusing on problems arising in continuous control, we propose a functional regularization approach to augmenting model-free RL. In particular, we regularize the behavior of the deep policy to be similar to a control prior, i.e., we regularize in function space. We show that functional regularization yields a bias-variance trade-off, and propose an adaptive tuning strategy to optimize this trade-off. When the prior policy has control-theoretic stability guarantees, we further show that this regularization approximately preserves those stability guarantees throughout learning. We validate our approach empirically on a wide range of settings, and demonstrate significantly reduced variance, guaranteed dynamic stability, and more efficient learning than deep RL alone. 

   more » « less
2. Leveraging Factored Action Spaces for Efficient Offline Reinforcement Learning in Healthcare

   Tang, Shengpu; Makar, Maggie; Sjoding, Michael; Doshi-Velez, Finale; Wiens, Jenna (January 2022, Advances in neural information processing systems)

   Many reinforcement learning (RL) applications have combinatorial action spaces, where each action is a composition of sub-actions. A standard RL approach ignores this inherent factorization structure, resulting in a potential failure to make meaningful inferences about rarely observed sub-action combinations; this is particularly problematic for offline settings, where data may be limited. In this work, we propose a form of linear Q-function decomposition induced by factored action spaces. We study the theoretical properties of our approach, identifying scenarios where it is guaranteed to lead to zero bias when used to approximate the Q-function. Outside the regimes with theoretical guarantees, we show that our approach can still be useful because it leads to better sample efficiency without necessarily sacrificing policy optimality, allowing us to achieve a better bias-variance trade-off. Across several offline RL problems using simulators and real-world datasets motivated by healthcare, we demonstrate that incorporating factored action spaces into valuebased RL can result in better-performing policies. Our approach can help an agent make more accurate inferences within underexplored regions of the state-action space when applying RL to observational datasets. 

   more » « less
3. Statistical Data Privacy: A Song of Privacy and Utility

   https://doi.org/10.1146/annurev-statistics-033121-112921  

   Slavković, Aleksandra; Seeman, Jeremy (March 2023, Annual Review of Statistics and Its Application)

   To quantify trade-offs between increasing demand for open data sharing and concerns about sensitive information disclosure, statistical data privacy (SDP) methodology analyzes data release mechanisms that sanitize outputs based on confidential data. Two dominant frameworks exist: statistical disclosure control (SDC) and the more recent differential privacy (DP). Despite framing differences, both SDC and DP share the same statistical problems at their core. For inference problems, either we may design optimal release mechanisms and associated estimators that satisfy bounds on disclosure risk measures, or we may adjust existing sanitized output to create new statistically valid and optimal estimators. Regardless of design or adjustment, in evaluating risk and utility, valid statistical inferences from mechanism outputs require uncertainty quantification that accounts for the effect of the sanitization mechanism that introduces bias and/or variance. In this review, we discuss the statistical foundations common to both SDC and DP, highlight major developments in SDP, and present exciting open research problems in private inference. 

   more » « less
4. On Characterizing the Trade-off in Invariant Representation Learning

   Bashir Sadeghi, Sepehr Dehdashtian (December 2022, Transactions on machine learning research)

   Many applications of representation learning, such as privacy preservation, algorithmic fairness, and domain adaptation, desire explicit control over semantic information being discarded. This goal is formulated as satisfying two objectives: maximizing utility for predicting a target attribute while simultaneously being invariant (independent) to a known semantic attribute. Solutions to invariant representation learning (IRepL) problems lead to a trade-off between utility and invariance when they are competing. While existing works study bounds on this trade-off, two questions remain outstanding: 1) What is the exact trade-off between utility and invariance? and 2) What are the encoders (mapping the data to a representation) that achieve the trade-off, and how can we estimate it from training data? This paper addresses these questions for IRepLs in reproducing kernel Hilbert spaces (RKHS)s. Under the assumption that the distribution of a low-dimensional projection of high-dimensional data is approximately normal, we derive a closed-form solution for the global optima of the underlying optimization problem for encoders in RKHSs. This yields closed formulae for a near-optimal trade-off, corresponding optimal representation dimensionality, and the corresponding encoder(s). We also numerically quantify the trade-off on representative problems and compare them to those achieved by baseline IRepL algorithms. 

   more » « less
5. Regression-based causal inference with factorial experiments: estimands, model specifications and design-based properties

   https://doi.org/10.1093/biomet/asab051  

   Zhao, Anqi; Ding, Peng (October 2021, Biometrika)

   Summary Factorial designs are widely used because of their ability to accommodate multiple factors simultaneously. Factor-based regression with main effects and some interactions is the dominant strategy for downstream analysis, delivering point estimators and standard errors simultaneously via one least-squares fit. Justification of these convenient estimators from the design-based perspective requires quantifying their sampling properties under the assignment mechanism while conditioning on the potential outcomes. To this end, we derive the sampling properties of the regression estimators under a wide range of specifications, and establish the appropriateness of the corresponding robust standard errors for Wald-type inference. The results help to clarify the causal interpretation of the coefficients in these factor-based regressions, and motivate the definition of general factorial effects to unify the definitions of factorial effects in various fields. We also quantify the bias-variance trade-off between the saturated and unsaturated regressions from the design-based perspective. 

   more » « less