Practical Adversarial Multivalid Conformal Prediction

by   Osbert Bastani, et al.

We give a simple, generic conformal prediction method for sequential prediction that achieves target empirical coverage guarantees against adversarially chosen data. It is computationally lightweight – comparable to split conformal prediction – but does not require having a held-out validation set, and so all data can be used for training models from which to derive a conformal score. It gives stronger than marginal coverage guarantees in two ways. First, it gives threshold calibrated prediction sets that have correct empirical coverage even conditional on the threshold used to form the prediction set from the conformal score. Second, the user can specify an arbitrary collection of subsets of the feature space – possibly intersecting – and the coverage guarantees also hold conditional on membership in each of these subsets. We call our algorithm MVP, short for MultiValid Prediction. We give both theory and an extensive set of empirical evaluations.


page 17

page 22

page 24


Split Localized Conformal Prediction

Conformal prediction is a simple and powerful tool that can quantify unc...

Approximate Conditional Coverage via Neural Model Approximations

Constructing reliable prediction sets is an obstacle for applications of...

The limits of distribution-free conditional predictive inference

We consider the problem of distribution-free predictive inference, with ...

Classification with Valid and Adaptive Coverage

Conformal inference, cross-validation+, and the jackknife+ are hold-out ...

Split Conformal Prediction for Dependent Data

Split conformal prediction is a popular tool to obtain predictive interv...

CD-split: efficient conformal regions in high dimensions

Conformal methods create prediction bands that control average coverage ...

Online Multivalid Learning: Means, Moments, and Prediction Intervals

We present a general, efficient technique for providing contextual predi...