Adaptive inference is a promising technique to improve the computational efficiency of deep models at test time. In contrast to static models which use the same computation graph for all instances, adaptive networks can dynamically adjust their structure conditioned on each input. While existing research on adaptive inference mainly focuses on designing more advanced architectures, this paper investigates how to train such networks more effectively. Specifically, we consider a typical adaptive deep network with multiple intermediate classifiers. We present three techniques to improve its training efficacy from two aspects: 1) a Gradient Equilibrium algorithm to resolve the conflict of learning of different classifiers; 2) an Inline Subnetwork Collaboration approach and a One-for-all Knowledge Distillation algorithm to enhance the collaboration among classifiers. On multiple datasets (CIFAR-10, CIFAR-100 and ImageNet), we show that the proposed approach consistently leads to further improved efficiency on top of state-of-the-art adaptive deep networks.
|Title of host publication||Proceedings - 2019 International Conference on Computer Vision, ICCV 2019|
|Number of pages||10|
|State||Published - Oct 2019|
|Event||17th IEEE/CVF International Conference on Computer Vision, ICCV 2019 - Seoul, Korea, Republic of|
Duration: Oct 27 2019 → Nov 2 2019
|Name||Proceedings of the IEEE International Conference on Computer Vision|
|Conference||17th IEEE/CVF International Conference on Computer Vision, ICCV 2019|
|Country/Territory||Korea, Republic of|
|Period||10/27/19 → 11/2/19|
Bibliographical noteFunding Information:
Acknowledgements. Gao Huang is supported in part by Beijing Academy of Artificial Intelligence under grant BAAI2019QN0106. Hao Li is supported in part by Ts-inghua University Initiative Scientific Research Program and Tsinghua Academic Fund for Undergraduate Overseas Studies. We thank Danlu Chen for helpful discussions.
© 2019 IEEE.
ASJC Scopus subject areas
- Computer Vision and Pattern Recognition