Neural Networks' Extrapolative Abilities
Pedro and Tim discuss the extrapolative capabilities of neural networks, exploring how they decompose input spaces and generalize beyond training ranges. They delve into the concept of hyperplanes providing valuable information in the extrapolative regime, shedding light on the surprising extrapolative performance of basic MLPs in high-dimensional spaces.In this clip
From this podcast
Machine Learning Street Talk (MLST)
#96 Prof. PEDRO DOMINGOS - There are no infinities, utility functions, neurosymbolic
Related Questions