1 paper
Juliusz Ziomek, George Whittle, Michael A. Osborne
In spite of their prevalence, the behaviour of Neural Networks when extrapolating far from the training distribution remains poorly understood, with existing results limited to spe…