## Center for Applied Mathematics Colloquium

We study the problem of estimating target parameters in nonparametric models with nuisance parameters. Replacing unknown nuisance parameters with nonparametric estimators, such as machine learning (ML) methods, can result in “plug-in bias.” Methods that avoid sub-optimal bias-variance trade-off by performing a debiasing step of the initial estimate rely on the influence function (IF) of the target parameter as input. However, deriving the IF requires specialized expertise and obstructs the adaptation of these methods by practitioners. We propose a novel way to debias which (i) simultaneously debiases many different target parameters, (ii) does not require the IF for implementation, and (iii) is computationally tractable. Building on the TMLE framework, our method updates an initial estimate with a regularized likelihood maximization step using a nonparametric reproducing kernel Hilbert space (RKHS)-based model. Our method can be readily adapted to new estimation problems and automated without analytic derivations, offering the efficiency of competing debiasing techniques while retaining the utility of the plug-in approach.