AlphaNet: Improved Training of Supernets with Alpha-Divergence

          Dilin Wang 1 Chengyue Gong * 2 Meng Li * 1 Qiang Liu 2 Vikas Chandra 1

            Abstract                Neural architecture search (NAS) automates the neural net-
   Weight-sharing neural architecture search (NAS)      work design by exploring an enormous architecture space
   is an effective technique for automating efficient    and achieves state-of-the-art (SOTA) performance on var-
   neural architecture design. We ...