Low Dimensional Trajectory Hypothesis is True: DNNs Can Be Trained in Tiny Subspaces
Sign Up to like & getrecommendations! Published in 2022 at "IEEE Transactions on Pattern Analysis and Machine Intelligence"
DOI: 10.1109/tpami.2022.3178101
Abstract: Deep neural networks (DNNs) usually contain massive parameters, but there is redundancy such that it is guessed that they could be trained in low-dimensional subspaces. In this paper, we propose a Dynamic Linear Dimensionality Reduction… read more here.
Keywords: true dnns; dimensional trajectory; trajectory hypothesis; dnns trained ... See more keywords