arXiv subjects

Antoine Chambaz

Publications and source records attributed to Antoine Chambaz.

At least 19 recordsLinked to original sources

arXiv2026

Set-Valued Policy Learning

Conventional treatment policies map patient covariates to a single recommended intervention in order to maximize expected clinical outcomes. Although a rich body of causal inference methods has been developed to estimate such policies, point-valued recommendations can be highly sensitive to estimation uncertainty, model specification, and finite-sample variability, while typically providing little guidance about how confident one should be in the recommended action. In this work, we propose a set-valued policy learning paradigm for the multiple-treatment setting, in which policies output a set of plausible treatments rather than a single recommendation. This formulation enables intrinsic uncertainty quantification, with the size of the predicted set reflecting the degree of decision ambiguity. We extend the learning-to-defer framework to multiple treatments via a novel \textit{greatest Lower Bound} method, and introduce \textit{conformal policy learning}, which bridges the gap between unobserved ground-truth optimal treatments and estimated optimal treatment rules. Drawing on insights from the noisy-label literature, we develop a randomness-injection approach that guarantees marginal coverage without requiring assumptions on underlying black-box optimal treatment rules. Through experiments on synthetic data and a real-world application to In-Vitro Fertilization (IVF), we demonstrate that our methods produce robust and actionable policies that naturally incorporate clinical considerations while effectively balancing performance and reliability.

cs.LG↗

arXiv2026

A Topological Sorting Criterion for Random Causal Directed Acyclic Graphs

Random directed acyclic graphs (DAGs) based on imposing an order on Erd\H{o}s-R\'enyi and scale free random graphs are widely used for evaluating causal discovery algorithms. We show that in such DAGs, the set of nodes reachable via open paths, termed relatives, increases monotonically along the causal order. We assess the prevalence of this pattern numerically, and demonstrate that it can be exploited for causal order recovery via sorting by the estimated number of relatives. We note that many simulations in the literature feature settings where this yields an excellent proxy for the causal order, and show that a strict increase of relatives along the causal order leads to a singular Markov equivalence class. We propose sampling time-series DAGs as a possible alternative and discuss implications for causal discovery algorithms and their evaluation on synthetic data.

stat.ME↗

arXiv2026

Policy learning under constraint: Maximizing a primary outcome while controlling an adverse event

A medical policy aims to support decision-making by mapping patient characteristics to individualized treatment recommendations. Standard approaches typically optimize a single outcome criterion. For example, recommending treatment according to the sign of the Conditional Average Treatment Effect (CATE) maximizes the policy "value" by exploiting treatment effect heterogeneity. This point of view shifts policy learning towards the challenge of learning a reliable CATE estimator. However, in multi-outcome settings, such strategies ignore the risk of adverse events, despite their relevance. PLUC (Policy Learning Under Constraint) addresses this challenges by learning an estimator of the CATE that yields smoothed policies controlling the probability of an adverse event in observational settings. Inspired by insights from EP-learning, PLUC involves the optimization of strongly convex Lagrangian criteria over a convex hull of functions. Its alternating procedure iteratively applies the Frank-Wolfe algorithm to minimize the current criterion, then performs a targeting step that updates the criterion so that its evaluations at previously visited landmarks become targeted estimators of the corresponding theoretical quantities. An R package PLUC-R provides a practical implementation. We illustrate PLUC's performance through a series of numerical experiments.

stat.ME↗

arXiv2025

Transforming Conditional Density Estimation Into a Single Nonparametric Regression Task

We propose a way of transforming the problem of conditional density estimation into a single nonparametric regression task via the introduction of auxiliary samples. This allows leveraging regression methods that work well in high dimensions, such as neural networks and decision trees. Our main theoretical result characterizes and establishes the convergence of our estimator to the true conditional density in the data limit. We develop condensité, a method that implements this approach. We demonstrate the benefit of the auxiliary samples on synthetic data and showcase that condensité can achieve good out-of-the-box results. We evaluate our method on a large population survey dataset and on a satellite imaging dataset. In both cases, we find that condensité matches or outperforms the state of the art and yields conditional densities in line with established findings in the literature on each dataset. Our contribution opens up new possibilities for regression-based conditional density estimation and the empirical results indicate strong promise for applied research.

stat.ML↗

arXiv2025

The Case for Time in Causal DAGs

We make the case for incorporating a notion of time into causal directed acyclic graphs (DAGs). We demonstrate that nontemporal causal DAGs are ambiguous and obstruct justification of the acyclicity assumption. Assuming that causes precede effects, causal relationships are relative to the time order, and causal DAGs require temporal qualification. We propose a formalization via composite causal variables that refer to quantities at one or multiple time points. We emphasize that the acyclicity assumption requires different justifications depending on whether the time order allows cycles. We conclude by discussing implications for the interpretation and applicability of DAGs as causal models.

stat.ME↗

arXiv2023

AdaptiveConformal: An R Package for Adaptive Conformal Inference

Conformal Inference (CI) is a popular approach for generating finite sample prediction intervals based on the output of any point prediction method when data are exchangeable. Adaptive Conformal Inference (ACI) algorithms extend CI to the case of sequentially observed data, such as time series, and exhibit strong theoretical guarantees without having to assume exchangeability of the observed data. The common thread that unites algorithms in the ACI family is that they adaptively adjust the width of the generated prediction intervals in response to the observed data. We provide a detailed description of five ACI algorithms and their theoretical guarantees, and test their performance in simulation studies. We then present a case study of producing prediction intervals for influenza incidence in the United States based on black-box point forecasts. Implementations of all the algorithms are released as an open-source R package, AdaptiveConformal, which also includes tools for visualizing and summarizing conformal prediction intervals.

stat.CO↗

arXiv2023

A Scale-Invariant Sorting Criterion to Find a Causal Order in Additive Noise Models

Additive Noise Models (ANMs) are a common model class for causal discovery from observational data and are often used to generate synthetic data for causal discovery benchmarking. Specifying an ANM requires choosing all parameters, including those not fixed by explicit assumptions. Reisach et al. (2021) show that sorting variables by increasing variance often yields an ordering close to a causal order and introduce var-sortability to quantify this alignment. Since increasing variances may be unrealistic and are scale-dependent, ANM data are often standardized in benchmarks. We show that synthetic ANM data are characterized by another pattern that is scale-invariant: the explainable fraction of a variable's variance, as captured by the coefficient of determination $R^2$, tends to increase along the causal order. The result is high $R^2$-sortability, meaning that sorting the variables by increasing $R^2$ yields an ordering close to a causal order. We propose an efficient baseline algorithm termed $R^2$-SortnRegress that exploits high $R^2$-sortability and that can match and exceed the performance of established causal discovery algorithms. We show analytically that sufficiently high edge weights lead to a relative decrease of the noise contributions along causal chains, resulting in increasingly deterministic relationships and high $R^2$. We characterize $R^2$-sortability for different simulation parameters and find high values in common settings. Our findings reveal high $R^2$-sortability as an assumption about the data generating process relevant to causal discovery and implicit in many ANM sampling schemes. It should be made explicit, as its prevalence in real-world data is unknown. For causal discovery benchmarking, we implement $R^2$-sortability, the $R^2$-SortnRegress algorithm, and ANM simulation procedures in our library CausalDisco at https://causaldisco.github.io/CausalDisco/.

stat.ML↗

arXiv2023

Quantile Super Learning for independent and online settings with application to solar power forecasting

Estimating quantiles of an outcome conditional on covariates is of fundamental interest in statistics with broad application in probabilistic prediction and forecasting. We propose an ensemble method for conditional quantile estimation, Quantile Super Learning, that combines predictions from multiple candidate algorithms based on their empirical performance measured with respect to a cross-validated empirical risk of the quantile loss function. We present theoretical guarantees for both iid and online data scenarios. The performance of our approach for quantile estimation and in forming prediction intervals is tested in simulation studies. Two case studies related to solar energy are used to illustrate Quantile Super Learning: in an iid setting, we predict the physical properties of perovskite materials for photovoltaic cells, and in an online setting we forecast ground solar irradiance based on output from dynamic weather ensemble models.

stat.ME↗

arXiv2023

Positivity-free Policy Learning with Observational Data

Policy learning utilizing observational data is pivotal across various domains, with the objective of learning the optimal treatment assignment policy while adhering to specific constraints such as fairness, budget, and simplicity. This study introduces a novel positivity-free (stochastic) policy learning framework designed to address the challenges posed by the impracticality of the positivity assumption in real-world scenarios. This framework leverages incremental propensity score policies to adjust propensity score values instead of assigning fixed values to treatments. We characterize these incremental propensity score policies and establish identification conditions, employing semiparametric efficiency theory to propose efficient estimators capable of achieving rapid convergence rates, even when integrated with advanced machine learning algorithms. This paper provides a thorough exploration of the theoretical guarantees associated with policy learning and validates the proposed framework's finite-sample performance through comprehensive numerical experiments, ensuring the identification of causal effects from observational data is both robust and reliable.

stat.ME↗

arXiv2023

Forecasting the cost of drought events in France by Super Learning

Drought events are the second most expensive type of natural disaster within the French legal framework known as the natural disasters compensation scheme. In recent years, drought events have been remarkable in their geographical scale and intensity. We develop and apply a new methodology to forecast the cost of a drought event in France. The methodology hinges on Super Learning (van der Laan et al., 2007; Benkeser et al., 2018), a general aggregation strategy to learn a feature of the law of the data identified through an ad hoc risk function by relying on a library of algorithms. The algorithms either compete (discrete Super Learning) or collaborate (continuous Super Learning), with a cross-validation scheme determining the best performing algorithm or combination of algorithms, respectively. Our Super Learner takes into account the complex dependence structure induced in the data by the spatial and temporal nature of drought events.

stat.AP↗

arXiv2023

Optimal transport-based machine learning to match specific patterns: application to the detection of molecular regulation patterns in omics data

We present several algorithms designed to learn a pattern of correspondence between two data sets in situations where it is desirable to match elements that exhibit a relationship belonging to a known parametric model. In the motivating case study, the challenge is to better understand micro-RNA regulation in the striatum of Huntington's disease model mice. The algorithms unfold in two stages. First, an optimal transport plan P and an optimal affine transformation are learned, using the Sinkhorn-Knopp algorithm and a mini-batch gradient descent. Second, P is exploited to derive either several co-clusters or several sets of matched elements. A simulation study illustrates how the algorithms work and perform. The real data application further illustrates their applicability and interest.

q-bio.GN↗

arXiv2023

Inference in Marginal Structural Models by Automatic Targeted Bayesian and Minimum Loss-Based Estimation

Two of the principle tasks of causal inference are to define and estimate the effect of a treatment on an outcome of interest. Formally, such treatment effects are defined as a possibly functional summary of the data generating distribution, and are referred to as target parameters. Estimation of the target parameter can be difficult, especially when it is high-dimensional. Marginal Structural Models (MSMs) provide a way to summarize such target parameters in terms of a lower dimensional working model. We introduce the semi-parametric efficiency bound for estimating MSM parameters in a general setting. We then present a frequentist estimator that achieves this bound based on Targeted Minimum Loss-Based Estimation. Our results are derived in a general context, and can be easily adapted to specific data structures and target parameters. We then describe a novel targeted Bayesian estimator and provide a Bernstein von-Mises type result analyzing its asymptotic behavior. We propose a universal algorithm that uses automatic differentiation to put the estimator into practice for arbitrary choice of working model. The frequentist and Bayesian estimators have been implemented in the Julia software package TargetedMSM.jl. Finally, we illustrate our proposed methods by investigating the effect of interventions on family planning behavior using data from a randomized field experiment conducted in Malawi.

math.ST↗

arXiv2021

Personalized Online Machine Learning

In this work, we introduce the Personalized Online Super Learner (POSL) -- an online ensembling algorithm for streaming data whose optimization procedure accommodates varying degrees of personalization. Namely, POSL optimizes predictions with respect to baseline covariates, so personalization can vary from completely individualized (i.e., optimization with respect to baseline covariate subject ID) to many individuals (i.e., optimization with respect to common baseline covariates). As an online algorithm, POSL learns in real-time. POSL can leverage a diversity of candidate algorithms, including online algorithms with different training and update times, fixed algorithms that are never updated during the procedure, pooled algorithms that learn from many individuals' time-series, and individualized algorithms that learn from within a single time-series. POSL's ensembling of this hybrid of base learning strategies depends on the amount of data collected, the stationarity of the time-series, and the mutual characteristics of a group of time-series. In essence, POSL decides whether to learn across samples, through time, or both, based on the underlying (unknown) structure in the data. For a wide range of simulations that reflect realistic forecasting scenarios, and in a medical data application, we examine the performance of POSL relative to other current ensembling and online learning methods. We show that POSL is able to provide reliable predictions for time-series data and adjust to changing data-generating environments. We further cultivate POSL's practicality by extending it to settings where time-series enter/exit dynamically over chronological time.

stat.ML↗

arXiv2021

One-step ahead sequential Super Learning from short times series of many slightly dependent data, and anticipating the cost of natural disasters

Suppose that we observe a short time series where each time-t-specific data-structure consists of many slightly dependent data indexed by a and that we want to estimate a feature of the law of the experiment that depends neither on t nor on a. We develop and study an algorithm to learn sequentially which base algorithm in a user-supplied collection best carries out the estimation task in terms of excess risk and oracular inequalities. The analysis, which uses dependency graph to model the amount of conditional independence within each t-specific data-structure and a concentration inequality by Janson [2004], leverages a large ratio of the number of distinct a's to the degree of the dependency graph in the face of a small number of t-specific data-structures. The so-called one-step ahead Super Learner is applied to the motivating example where the challenge is to anticipate the cost of natural disasters in France.

math.ST↗

arXiv2021

Optimal tests of the composite null hypothesis arising in mediation analysis

The indirect effect of an exposure on an outcome through an intermediate variable can be identified by a product of two regression coefficients under certain causal and regression modeling assumptions. In this context, the null hypothesis of no indirect effect is a composite null hypothesis, as the null holds if either regression coefficient is zero. A consequence is that traditional hypothesis tests are severely underpowered near the origin (i.e., when both coefficients are small with respect to standard errors). We propose hypothesis tests that (i) preserve level alpha type 1 error, (ii) meaningfully improve power when both true underlying effects are small relative to sample size, and (iii) preserve power when at least one is not. One approach gives a closed-form test that is minimax optimal with respect to local power over the alternative parameter space. Another uses sparse linear programming to produce an approximately optimal test for a Bayes risk criterion. We discuss adaptations for performing large-scale hypothesis testing as well as modifications that yield improved interpretability. We provide an R package that implements our proposed methodology.

math.ST↗

arXiv2021

Risk Minimization from Adaptively Collected Data: Guarantees for Supervised and Policy Learning

Empirical risk minimization (ERM) is the workhorse of machine learning, whether for classification and regression or for off-policy policy learning, but its model-agnostic guarantees can fail when we use adaptively collected data, such as the result of running a contextual bandit algorithm. We study a generic importance sampling weighted ERM algorithm for using adaptively collected data to minimize the average of a loss function over a hypothesis class and provide first-of-their-kind generalization guarantees and fast convergence rates. Our results are based on a new maximal inequality that carefully leverages the importance sampling structure to obtain rates with the right dependence on the exploration rate in the data. For regression, we provide fast rates that leverage the strong convexity of squared-error loss. For policy learning, we provide rate-optimal regret guarantees that close an open gap in the existing literature whenever exploration decays to zero, as is the case for bandit-collected data. An empirical investigation validates our theory.

stat.ML↗

arXiv2021

Post-Contextual-Bandit Inference

Contextual bandit algorithms are increasingly replacing non-adaptive A/B tests in e-commerce, healthcare, and policymaking because they can both improve outcomes for study participants and increase the chance of identifying good or even best policies. To support credible inference on novel interventions at the end of the study, nonetheless, we still want to construct valid confidence intervals on average treatment effects, subgroup effects, or value of new policies. The adaptive nature of the data collected by contextual bandit algorithms, however, makes this difficult: standard estimators are no longer asymptotically normally distributed and classic confidence intervals fail to provide correct coverage. While this has been addressed in non-contextual settings by using stabilized estimators, the contextual setting poses unique challenges that we tackle for the first time in this paper. We propose the Contextual Adaptive Doubly Robust (CADR) estimator, the first estimator for policy value that is asymptotically normal under contextual adaptive data collection. The main technical challenge in constructing CADR is designing adaptive and consistent conditional standard deviation estimators for stabilization. Extensive numerical experiments using 57 OpenML datasets demonstrate that confidence intervals based on CADR uniquely provide correct coverage.

stat.ML↗

arXiv2020

Rate-adaptive model selection over a collection of black-box contextual bandit algorithms

We consider the model selection task in the stochastic contextual bandit setting. Suppose we are given a collection of base contextual bandit algorithms. We provide a master algorithm that combines them and achieves the same performance, up to constants, as the best base algorithm would, if it had been run on its own. Our approach only requires that each algorithm satisfy a high probability regret bound. Our procedure is very simple and essentially does the following: for a well chosen sequence of probabilities $(p_{t})_{t\geq 1}$, at each round $t$, it either chooses at random which candidate to follow (with probability $p_{t}$) or compares, at the same internal sample size for each candidate, the cumulative reward of each, and selects the one that wins the comparison (with probability $1-p_{t}$). To the best of our knowledge, our proposal is the first one to be rate-adaptive for a collection of general black-box contextual bandit algorithms: it achieves the same regret rate as the best candidate. We demonstrate the effectiveness of our method with simulation studies.

cs.LG↗