Scipediacontent: Scipediacontent moved page Draft Content 697304703 to Xie et al 2020a

2021-02-02T05:17:54Z

Scipediacontent moved page Draft Content 697304703 to Xie et al 2020a

Scipediacontent: Created page with " == Abstract == utomatic designing computationally efficient neural networks has received much attention in recent years. Existing approaches either utilize network pruning o..."

2021-02-02T05:17:47Z

Created page with " == Abstract == utomatic designing computationally efficient neural networks has received much attention in recent years. Existing approaches either utilize network pruning o..."

New page

== Abstract ==

utomatic designing computationally efficient neural networks has received much attention in recent years. Existing approaches either utilize network pruning or leverage the network architecture search methods. This paper presents a new framework named network adjustment, which considers network accuracy as a function of FLOPs, so that under each network configuration, one can estimate the FLOPs utilization ratio (FUR) for each layer and use it to determine whether to increase or decrease the number of channels on the layer. Note that FUR, like the gradient of a non-linear function, is accurate only in a small neighborhood of the current network. Hence, we design an iterative mechanism so that the initial network undergoes a number of steps, each of which has a small 'adjusting rate' to control the changes to the network. The computational overhead of the entire search process is reasonable, i.e., comparable to that of re-training the final model from scratch. Experiments on standard image classification datasets and a wide range of base networks demonstrate the effectiveness of our approach, which consistently outperforms the pruning counterpart. The code is available at https://github.com/danczs/NetworkAdjustment.

== Original document ==

The different versions of the original document can be found in:

* [http://arxiv.org/abs/2004.02767 http://arxiv.org/abs/2004.02767]

* [http://arxiv.org/pdf/2004.02767 http://arxiv.org/pdf/2004.02767]

* [http://xplorestaging.ieee.org/ielx7/9142308/9156271/09156620.pdf?arnumber=9156620 http://xplorestaging.ieee.org/ielx7/9142308/9156271/09156620.pdf?arnumber=9156620],
: [http://dx.doi.org/10.1109/cvpr42600.2020.01067 http://dx.doi.org/10.1109/cvpr42600.2020.01067]

* [https://dblp.uni-trier.de/db/conf/cvpr/cvpr2020.html#Chen0XLWT20 https://dblp.uni-trier.de/db/conf/cvpr/cvpr2020.html#Chen0XLWT20],
: [https://openaccess.thecvf.com/content_CVPR_2020/papers/Chen_Network_Adjustment_Channel_Search_Guided_by_FLOPs_Utilization_Ratio_CVPR_2020_paper.pdf https://openaccess.thecvf.com/content_CVPR_2020/papers/Chen_Network_Adjustment_Channel_Search_Guided_by_FLOPs_Utilization_Ratio_CVPR_2020_paper.pdf],
: [https://openaccess.thecvf.com/content_CVPR_2020/html/Chen_Network_Adjustment_Channel_Search_Guided_by_FLOPs_Utilization_Ratio_CVPR_2020_paper.html https://openaccess.thecvf.com/content_CVPR_2020/html/Chen_Network_Adjustment_Channel_Search_Guided_by_FLOPs_Utilization_Ratio_CVPR_2020_paper.html],
: [https://ieeexplore.ieee.org/document/9156620 https://ieeexplore.ieee.org/document/9156620],
: [https://academic.microsoft.com/#/detail/3035422913 https://academic.microsoft.com/#/detail/3035422913]

← Older revision	Revision as of 05:17, 2 February 2021
(No difference)

Xie et al 2020a - Revision history

Scipediacontent: Scipediacontent moved page Draft Content 697304703 to Xie et al 2020a

Scipediacontent: Created page with " == Abstract == utomatic designing computationally efficient neural networks has received much attention in recent years. Existing approaches either utilize network pruning o..."