Our new X account is live! Follow @wizwand_team for updates
WorkDL logo mark

Learning to Defer to Multiple Experts: Consistent Surrogate Losses, Confidence Calibration, and Conformal Ensembles

About

We study the statistical properties of learning to defer (L2D) to multiple experts. In particular, we address the open problems of deriving a consistent surrogate loss, confidence calibration, and principled ensembling of experts. Firstly, we derive two consistent surrogates -- one based on a softmax parameterization, the other on a one-vs-all (OvA) parameterization -- that are analogous to the single expert losses proposed by Mozannar and Sontag (2020) and Verma and Nalisnick (2022), respectively. We then study the frameworks' ability to estimate P( m_j = y | x ), the probability that the jth expert will correctly predict the label for x. Theory shows the softmax-based loss causes mis-calibration to propagate between the estimates while the OvA-based loss does not (though in practice, we find there are trade offs). Lastly, we propose a conformal inference technique that chooses a subset of experts to query when the system defers. We perform empirical validation on tasks for galaxy, skin lesion, and hate speech classification.

Rajeev Verma, Daniel Barrej\'on, Eric Nalisnick• 2022

Related benchmarks

TaskDatasetResultRank
Learning to DeferImageNet Overlapped Dog Expert
Error Rate41.56
20
Learning to DeferCIFAR-100 Overlapped Animal Expert
Error Rate16.1
20
Learning to DeferCIFAR-100 Animal Expert
Error Rate18.48
20
Learning to DeferCIFAR-100 varying-accuracy synthetic expert (test)
Error Rate18.96
20
Learning to DeferImageNet Dog Expert
Error Rate42.74
20
Learning to DeferImageNet varying-accuracy synthetic expert (val)
Error42.58
20
ClassificationChaoyang
Error Rate1.32
16
ClassificationMiceBone
Error Rate15.17
16
Showing 8 of 8 rows

Other info

Follow for update