Asymptotically Optimal Regret for Black-Box Predict-then-Optimize

We consider the predict-then-optimize paradigm for decision-making in which a practitioner (1) trains a supervised learning model on historical data of decisions, contexts, and rewards, and then (2) uses the resulting model to make future binary decisions for new contexts by finding the decision that maximizes the model's predicted reward. This approach is common in industry. Past analysis assumes that rewards are observed for all actions for all historical contexts, which is possible only in problems with special structure. Motivated by problems from ads targeting and recommender systems, we study new black-box predict-then-optimize problems that lack this special structure and where we only observe the reward from the action taken. We present a novel loss function, which we call Empirical Soft Regret (ESR), designed to significantly improve reward when used in training compared to classical accuracy-based metrics like mean-squared error. This loss function targets the regret achieved when taking a suboptimal decision; because the regret is generally not differentiable, we propose a differentiable "soft" regret term that allows the use of neural networks and other flexible machine learning models dependent on gradient-based training. In the particular case of paired data, we show theoretically that optimizing our loss function yields asymptotically optimal regret within the class of supervised learning models. We also show our approach significantly outperforms state-of-the-art algorithms on real-world decision-making problems in news recommendation and personalized healthcare compared to benchmark methods from contextual bandits and conditional average treatment effect estimation.

Re-embedding data to strengthen recovery guarantees of clustering

We propose a clustering method that involves chaining four known techniques into a pipeline yielding an algorithm with stronger recovery guarantees than any of the four components separately. Given $n$ points in $\mathbb R^d$, the first component of our pipeline, which we call leapfrog distances, is reminiscent of density-based clustering, yielding an $n\times n$ distance matrix. The leapfrog distances are then translated to new embeddings using multidimensional scaling and spectral methods, two other known techniques, yielding new embeddings of the $n$ points in $\mathbb R^{d'}$, where $d'$ satisfies $d'\ll d$ in general. Finally, sum-of-norms (SON) clustering is applied to the re-embedded points. Although the fourth step (SON clustering) can in principle be replaced by any other clustering method, our focus is on provable guarantees of recovery of underlying structure. Therefore, we establish that the re-embedding improves recovery SON clustering, since SON clustering is a well-studied method that already has provable guarantees.

Regret Bounds and Experimental Design for Estimate-then-Optimize

In practical applications, data is used to make decisions in two steps: estimation and optimization. First, a machine learning model estimates parameters for a structural model relating decisions to outcomes. Second, a decision is chosen to optimize the structural model's predicted outcome as if its parameters were correctly estimated. Due to its flexibility and simple implementation, this ``estimate-then-optimize'' approach is often used for data-driven decision-making. Errors in the estimation step can lead estimate-then-optimize to sub-optimal decisions that result in regret, i.e., a difference in value between the decision made and the best decision available with knowledge of the structural model's parameters. We provide a novel bound on this regret for smooth and unconstrained optimization problems. Using this bound, in settings where estimated parameters are linear transformations of sub-Gaussian random vectors, we provide a general procedure for experimental design to minimize the regret resulting from estimate-then-optimize. We demonstrate our approach on simple examples and a pandemic control application.