arXiv research

New method improves robustness of double robust estimators under complete misspecification.

New methods combine machine learning with doubly robust estimators for better treatment effect estimation.

New algorithm reduces regret in GLM bandits with tighter bounds.

When training a machine learning model with observational data, it is often encountered that some values are systemically missing. Learning from the incomplete data in which the missingness depends on some covariates may lead to biased estimation of parameters and even harm the fairness of decision outcome. This paper …

New method for estimating parameters in inverse problems using double robustness.

New estimator improves ATT estimation efficiency with external controls.

Improved estimators for causal inference using cross-fitting and undersmoothing.

StableDR stabilizes doubly robust learning for biased recommendation data.

This paper investigates robust and efficient DR/RDR estimators for WATEs.

The paper shows some Montesinos links can't be doubly sliced strongly.

Paper introduces GDR-learners for estimating potential outcomes from observational data.

We develop a theory of chain complex double-cobordism for chain complexes equipped with Poincaré duality. The resulting double-cobordism groups are a refinement of Ranicki's torsion algebraic $L$-groups for localisations of a commutative ring with involution. The refinement is analogous to the difference between metabo…

A new algorithm improves offline reinforcement learning robustness.

A new method improves estimation of COVID-19 vaccine effectiveness.

Extends robust methods for causal inference, improving estimator performance.

Paper develops a new estimator for dynamic treatment effects in high-dimensional settings.

Sharp bounds on ATE with unmeasured confounders, valid even when misspecified.

Proposes a method to stabilize treatment effect estimation with unbalanced data.

Paper proposes methods to reduce bias and variance in recommender systems.

Proposes a robust method for predicting missing outcomes in covariate shift adaptation.

A knot in the three-sphere is doubly slice if it is the cross-section of an unknotted two-sphere in the four-sphere. For low-crossing knots, the most complete work to date gives a classification of doubly slice knots through 9 crossings. We extend that work through 12 crossings, resolving all but four cases among the 2…

We show that if the connected sum of two knots with coprime Alexander polynomials is doubly slice, then the Ozsváth-Szabó correction terms as smooth double sliceness obstructions vanish for both knots. Recently, Jeffrey Meier gave smoothly slice knots that are topologically doubly slice, but not smoothly doubly slice. …

Study proposes new OPE estimators for two-player zero-sum games.

Simplified tutorial on doubly robust learning for causal inference.

We develop new algebraic methods refining the Witt group of linking forms and Ranicki's torsion algebraic L-groups into double Witt groups and double L-groups. At each prime ideal of the underlying ring, our double Witt groups capture infinitely many more integral signatures of the linking form than the single Witt gro…

Proposes DR-ACI for causal effect intervals with temporal dependence.

Proposes a new method using GANs for testing conditional independence.

Study identifies prime strongly positive amphicheiral knots with double symmetry.

We propose a semiparametric test to evaluate (i) whether different instruments induce subpopulations of compliers with the same observable characteristics on average, and (ii) whether compliers have observable characteristics that are the same as the full population on average. The test is a flexible robustness check f…

We construct an infinite family of smoothly slice knots that we prove are topologically doubly slice. Using the correction terms coming from Heegaard Floer homology, we show that none of these knots is smoothly doubly slice. We use these knots to show that the subgroup of the double concordance group consisting of smoo…

A criterion ensures double sliceness for certain knots and satellite knots.

Tests validity of DML estimators without assumptions.

Develops a method to estimate optimal policy value in online learning.

For $\ell >1$, we develop $L^{(2)}$-signature obstructions for $(4\ell-3)$-dimensional knots with metabelian knot groups to be doubly slice. For each $\ell>1$, we construct an infinite family of knots on which our obstructions are non-zero, but for which double sliceness is not obstructed by any previously known invari…

Proposes CCME framework for estimating heterogeneous treatment effects.

Corrects mismatch in consistency of nuisance estimators for doubly robust methods.

The difference between slice and doubly-slice knots is reflected in algebra by the difference between metabolic and hyperbolic Blanchfield linking forms. We exploit this algebraic distinction to refine the classical Witt group of linking forms by defining a `double Witt group' of linking forms. We calculate the double …

The paper extends knot theory to 4-manifolds, defining new genera and obstructions.

Off-policy evaluation (OPE) in reinforcement learning allows one to evaluate novel decision policies without needing to conduct exploration, which is often costly or otherwise infeasible. We consider for the first time the semiparametric efficiency limits of OPE in Markov decision processes (MDPs), where actions, rewar…

Bayesian method corrects bias in treatment effect estimation.

We study the problem of off-policy critic evaluation in several variants of value-based off-policy actor-critic algorithms. Off-policy actor-critic algorithms require an off-policy critic evaluation step, to estimate the value of the new policy after every policy gradient update. Despite enormous success of off-policy …

Doubly robust self-training improves semi-supervised learning by balancing labeled and pseudo-labeled data.

While model selection is a well-studied topic in parametric and nonparametric regression or density estimation, selection of possibly high-dimensional nuisance parameters in semiparametric problems is far less developed. In this paper, we propose a selective machine learning framework for making inferences about a fini…

New estimators improve causal inference in machine learning studies.

Improves robustness of propensity score estimators in challenging settings.

The consistency of doubly robust estimators relies on consistent estimation of at least one of two nuisance regression parameters. In moderate to large dimensions, the use of flexible data-adaptive regression estimators may aid in achieving this consistency. However, $n^{1/2}$-consistency of doubly robust estimators is…

Novel characterization of augmented balancing weights combining outcome and weighting models.

Doubly-stochastic normalization improves robustness to heteroskedastic noise.