Search papers, labs, and topics across Lattice.
This paper introduces Wasserstein Filtering (WF), a sample selection method designed to recover clean population distributions from contaminated datasets by discarding suspicious samples based on their Wasserstein distance from the contaminated empirical distribution. The authors develop three computationally efficient algorithms鈥擲inkMarg, SinkWF, and SlicedWF鈥攖hat leverage optimal transport techniques to isolate and remove outliers effectively. Theoretical analysis under the Far Exclusion and Local Projection contamination model shows that WF achieves minimax optimality for distribution families with bounded covariance, while empirical results demonstrate its efficacy in outlier detection and generative modeling tasks.
Wasserstein Filtering can significantly enhance the robustness of generative models by effectively isolating and removing outliers from contaminated datasets.
Given a dataset where a portion of the samples are contaminated, our goal is to recover the underlying clean population distribution. To this end, we propose Wasserstein Filtering (WF), a novel sample selection framework that discards a fraction of suspicious samples and estimates the target distribution using the empirical measure of the remaining data. The core insight is to select a subset of samples whose empirical distribution maximizes its Wasserstein distance to the fully contaminated empirical distribution, thereby preferentially isolating and removing geometrically influential outliers. To render this optimization computationally tractable, we introduce three algorithms: a marginal screening scheme, SinkMarg, and two joint optimization algorithms, SinkWF and SlicedWF, leveraging entropic optimal transport and sliced Wasserstein approximations, respectively. On the theoretical front, we introduce the Far Exclusion and Local Projection (FELP) contamination model, which characterizes corruptions consisting of well-separated outliers and locally indistinguishable perturbations. Under this model, we prove that the WF estimator achieves minimax optimality over distribution families with bounded covariance. Extensive numerical experiments on synthetic datasets, benchmark anomaly detection suites, and robust generative learning with diffusion models demonstrate that WF serves as a highly practical, model-agnostic preprocessing tool. It delivers competitive outlier detection performance and provides substantial downstream benefits for generative modeling under heavy contamination.
