no code implementations • 1 Jan 2021 • Teresa Yeo, Oguzhan Fatih Kar, Amir Zamir
We present a method for making predictions using neural networks that, at the test time, is robust against shifts from the training data distribution.