Timezone: »

Practical Adversarial Multivalid Conformal Prediction
Osbert Bastani · Varun Gupta · Christopher Jung · Georgy Noarov · Ramya Ramalingam · Aaron Roth

Wed Nov 30 09:00 AM -- 11:00 AM (PST) @ Hall J #810

We give a simple, generic conformal prediction method for sequential prediction that achieves target empirical coverage guarantees on adversarial 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. Furthermore, 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 will 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.

Author Information

Osbert Bastani (University of Pennsylvania)
Varun Gupta (School of Engineering and Applied Science, University of Pennsylvania)
Christopher Jung (University of Pennsylvania)
Georgy Noarov (School of Engineering and Applied Science, University of Pennsylvania)
Ramya Ramalingam (University of Pennsylvania, University of Pennsylvania)
Aaron Roth (University of Pennsylvania)

More from the Same Authors