We therefore explore training amortized models with noisy labels, and we find that this is inexpensive and surprisingly effective.

It contains the entire family of hyperboloids as submanifolds.

We develop a method to generate prediction intervals that have a user-specified coverage level across all regions of feature-space, a property calledconditional coverage.

Existingmethods assume that the examples illustrating a concept are mapped in a fixed direction oftheDNN'slatent space.