NeurIPS 2019
Sun Dec 8th through Sat the 14th, 2019 at Vancouver Convention Center
Paper ID:8301
Title:Likelihood Ratios for Out-of-Distribution Detection

This paper applies a likelihood ratio test to detect out of distribution data, with several comparisons, and a new genomics dataset. Reviewers are supportive of this paper, and appreciated the author feedback. While ensembles perform slightly better in the CIFAR-10 vs SVHN experiment, these results will be important to include in revisions, for a balanced presentation. Great work!