Efficient Forward Architecture Search
ADVANCES IN NEURAL INFORMATION PROCESSING SYSTEMS 32 (NIPS 2019), pp. 10122-10131, 2019.
We propose a neural architecture search (NAS) algorithm, Petridish, to iteratively add shortcut connections to existing network layers. The added shortcut connections effectively perform gradient boosting on the augmented layers. The proposed algorithm is motivated by the feature selection algorithm forward stage-wise linear regression, s...More
PPT (Upload PPT)