1 paper · 1 filter
Sahil Singla, Atoosa Malemir Chegini, Mazda Moayeri +1
Deep neural networks can be unreliable in the real world when the training set does not adequately cover all the settings where they are deployed. Focusing on image classification,…