While the research on convolutional neural networks (CNNs) is progressing
quickly, the real-world deployment of these models is often limited by
computing resources and memory constraints. In this paper, we address this
issue by proposing a novel filter pruning method to compress and accelerate
CNNs. Our work is based on the linear relationship identified in different
feature map subspaces via visualization of feature maps. Such linear
relationship implies that the information in CNNs is redundant. Our method
eliminates the redundancy in convolutional filters by applying subspace
clustering to feature maps. In this way, most of the representative information
in the network can be retained in each cluster. Therefore, our method provides
an effective solution to filter pruning for which most existing methods
directly remove filters based on simple heuristics. The proposed method is
independent of the network structure, thus it can be adopted by any
off-the-shelf deep learning libraries. Experiments on different networks and
tasks show that our method outperforms existing techniques before fine-tuning,
and achieves the state-of-the-art results after fine-tuning.
Description
[1803.05729] Exploring Linear Relationship in Feature Map Subspace for ConvNets Compression
%0 Generic
%1 wang2018exploring
%A Wang, Dong
%A Zhou, Lei
%A Zhang, Xueni
%A Bai, Xiao
%A Zhou, Jun
%D 2018
%K cnn compression
%T Exploring Linear Relationship in Feature Map Subspace for ConvNets
Compression
%U http://arxiv.org/abs/1803.05729
%X While the research on convolutional neural networks (CNNs) is progressing
quickly, the real-world deployment of these models is often limited by
computing resources and memory constraints. In this paper, we address this
issue by proposing a novel filter pruning method to compress and accelerate
CNNs. Our work is based on the linear relationship identified in different
feature map subspaces via visualization of feature maps. Such linear
relationship implies that the information in CNNs is redundant. Our method
eliminates the redundancy in convolutional filters by applying subspace
clustering to feature maps. In this way, most of the representative information
in the network can be retained in each cluster. Therefore, our method provides
an effective solution to filter pruning for which most existing methods
directly remove filters based on simple heuristics. The proposed method is
independent of the network structure, thus it can be adopted by any
off-the-shelf deep learning libraries. Experiments on different networks and
tasks show that our method outperforms existing techniques before fine-tuning,
and achieves the state-of-the-art results after fine-tuning.
@misc{wang2018exploring,
abstract = {While the research on convolutional neural networks (CNNs) is progressing
quickly, the real-world deployment of these models is often limited by
computing resources and memory constraints. In this paper, we address this
issue by proposing a novel filter pruning method to compress and accelerate
CNNs. Our work is based on the linear relationship identified in different
feature map subspaces via visualization of feature maps. Such linear
relationship implies that the information in CNNs is redundant. Our method
eliminates the redundancy in convolutional filters by applying subspace
clustering to feature maps. In this way, most of the representative information
in the network can be retained in each cluster. Therefore, our method provides
an effective solution to filter pruning for which most existing methods
directly remove filters based on simple heuristics. The proposed method is
independent of the network structure, thus it can be adopted by any
off-the-shelf deep learning libraries. Experiments on different networks and
tasks show that our method outperforms existing techniques before fine-tuning,
and achieves the state-of-the-art results after fine-tuning.},
added-at = {2018-03-16T15:00:58.000+0100},
author = {Wang, Dong and Zhou, Lei and Zhang, Xueni and Bai, Xiao and Zhou, Jun},
biburl = {https://www.bibsonomy.org/bibtex/2dcc525683fd83c547f20df1454664a73/rcb},
description = {[1803.05729] Exploring Linear Relationship in Feature Map Subspace for ConvNets Compression},
interhash = {6f96988c6dbacf2cdd85a0b28d8eb4c4},
intrahash = {dcc525683fd83c547f20df1454664a73},
keywords = {cnn compression},
note = {cite arxiv:1803.05729Comment: 17 pages, 4 figures},
timestamp = {2018-03-16T15:00:58.000+0100},
title = {Exploring Linear Relationship in Feature Map Subspace for ConvNets
Compression},
url = {http://arxiv.org/abs/1803.05729},
year = 2018
}