Differentiable Neural Architecture Search in Equivalent Space with Exploration Enhancement

Abstract

Recent works on One-Shot Neural Architecture Search (NAS) mostly adopt a bilevel optimization scheme to alternatively optimize the supernet weights and architecture parameters after relaxing the discrete search space into a differentiable space. However, the non-negligible incongruence in their relaxation methods is hard to guarantee the differentiable optimization in the continuous space is equivalent to the optimization in the discrete space. Differently, this paper utilizes a variational graph autoencoder to injectively transform the discrete architecture space into an equivalently continuous latent space, to resolve the incongruence. A probabilistic exploration enhancement method is accordingly devised to encourage intelligent exploration during the architecture search in the latent space, to avoid local optimal in architecture search. As the catastrophic forgetting in differentiable One-Shot NAS deteriorates supernet predictive ability and makes the bilevel optimization inefficient, this paper further proposes an architecture complementation method to relieve this deficiency. We analyze the effectiveness of the proposed method, and a series of experiments have been conducted to compare the proposed method with state-of-the-art One-Shot NAS methods.

Cite

Text

Zhang et al. "Differentiable Neural Architecture Search in Equivalent Space with Exploration Enhancement." Neural Information Processing Systems, 2020.

Markdown

[Zhang et al. "Differentiable Neural Architecture Search in Equivalent Space with Exploration Enhancement." Neural Information Processing Systems, 2020.](https://mlanthology.org/neurips/2020/zhang2020neurips-differentiable/)

BibTeX

@inproceedings{zhang2020neurips-differentiable,
  title     = {{Differentiable Neural Architecture Search in Equivalent Space with Exploration Enhancement}},
  author    = {Zhang, Miao and Li, Huiqi and Pan, Shirui and Chang, Xiaojun and Ge, Zongyuan and Su, Steven},
  booktitle = {Neural Information Processing Systems},
  year      = {2020},
  url       = {https://mlanthology.org/neurips/2020/zhang2020neurips-differentiable/}
}