Sublinear Algorithms for Wasserstein and Total Variation Distances: Applications to Fairness and Privacy Auditing
This work addresses the need for resource-efficient distance estimation in streaming data, which is incremental but extends mergeable summaries to continuous distributions for practical use in fairness and privacy auditing.
The paper tackles the problem of efficiently computing Wasserstein and Total Variation distances between probability distributions from streaming samples, proposing a framework that uses sublinear space and improves on super-linear time and linear space methods, with tight results relative to lower bounds and applications in fairness and privacy auditing.
Resource-efficiently computing representations of probability distributions and the distances between them while only having access to the samples is a fundamental and useful problem across mathematical sciences. In this paper, we propose a generic framework to learn the probability and cumulative distribution functions (PDFs and CDFs) of a sub-Weibull, i.e. almost any light- or heavy-tailed, distribution while the samples from it arrive in a stream. The idea is to reduce these problems into estimating the frequency of an \textit{appropriately chosen subset} of the support of a \textit{properly discretised distribution}. We leverage this reduction to compute mergeable summaries of distributions from the stream of samples while requiring only sublinear space relative to the number of observed samples. This allows us to estimate Wasserstein and Total Variation (TV) distances between any two distributions while samples arrive in streams and from multiple sources. Our algorithms significantly improves on the existing methods for distance estimation incurring super-linear time and linear space complexities, and further extend the mergeable summaries framework to continuous distributions with possibly infinite support. Our results are tight with respect to the existing lower bounds for bounded discrete distributions. In addition, we leverage our proposed estimators of Wasserstein and TV distances to tightly audit the fairness and privacy of algorithms. We empirically demonstrate the efficiency of proposed algorithms across synthetic and real-world datasets.