Exploiting Operation Importance for Differentiable Neural Architecture Search
Sign Up to like & getrecommendations! Published in 2021 at "IEEE transactions on neural networks and learning systems"
DOI: 10.1109/tnnls.2021.3072950
Abstract: Recently, differentiable neural architecture search (NAS) methods have made significant progress in reducing the computational costs of NASs. Existing methods search for the best architecture by choosing candidate operations with higher architecture weights. However, architecture… read more here.
Keywords: differentiable neural; operation importance; architecture; search ... See more keywords