Timezone: »

DATA: Differentiable ArchiTecture Approximation
Jianlong Chang · xinbang zhang · Yiwen Guo · GAOFENG MENG · SHIMING XIANG · Chunhong Pan

Tue Dec 10 05:30 PM -- 07:30 PM (PST) @ East Exhibition Hall B + C #2

Neural architecture search (NAS) is inherently subject to the gap of architectures during searching and validating. To bridge this gap, we develop Differentiable ArchiTecture Approximation (DATA) with an Ensemble Gumbel-Softmax (EGS) estimator to automatically approximate architectures during searching and validating in a differentiable manner. Technically, the EGS estimator consists of a group of Gumbel-Softmax estimators, which is capable of converting probability vectors to binary codes and passing gradients from binary codes to probability vectors. Benefiting from such modeling, in searching, architecture parameters and network weights in the NAS model can be jointly optimized with the standard back-propagation, yielding an end-to-end learning mechanism for searching deep models in a large enough search space. Conclusively, during validating, a high-performance architecture that approaches to the learned one during searching is readily built. Extensive experiments on a variety of popular datasets strongly evidence that our method is capable of discovering high-performance architectures for image classification, language modeling and semantic segmentation, while guaranteeing the requisite efficiency during searching.

Author Information

Jianlong Chang (National Laboratory of Pattern Recognition, Institute of Automation, Chinese Academy of Sciences)
xinbang zhang (Institute of Automation,Chinese Academy of Science)
Yiwen Guo (Bytedance AI Lab)
GAOFENG MENG (Institute of Automation, Chinese Academy of Sciences)
SHIMING XIANG (Chinese Academy of Sciences, China)
Chunhong Pan (Institute of Automation, Chinese Academy of Sciences)

More from the Same Authors