Accelerating Convolutional Neural Networks for Mobile Applications

By Peisong Wang, Jian Cheng. 2016

75 citations recorded by the source corpus.

Convolutional neural networks (CNNs) have achieved remarkable performance in a wide range of computer vision tasks, typically at the cost of massive computational complexity. The low speed of these networks may hinder real-time applications especially when computational resources are limited. In this paper, an efficient and effective approach is proposed to accelerate the test-phase computation of CNNs based on low-rank and group sparse tensor decomposition. Specifically, for each convolutional layer, the kernel tensor is decomposed into the sum of a small number of low multilinear rank tensors. Then we replace the original kernel tensors in all layers with the approximate tensors and fine-tune the whole net with respect to the final classification task using standard backpropagation. \\ Comprehensive experiments on ILSVRC-12 demonstrate significant reduction in computational complexity, at the cost of negligible loss in accuracy. For the widely used VGG-16 model, our approach obtains a 6.6$\times$ speed-up on PC and 5.91$\times$ speed-up on mobile device of the whole network with less than 1\% increase on top-5 error.

Subfields: Computational Mathematics, Computer Vision and Pattern Recognition

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Computational Mathematics

In Computer Vision and Pattern Recognition

Citations in the corpus

What is the State of Neural Network Pruning? (2020, background): “Even when reporting FLOPs, which is nominally a consistent metric, different authors measure it differently (e.g., (Molchanov et al., 2016) vs (Wang & Cheng, 2016)), though most often papers entirely omit their formula for computing FLOPs.”
Recent Advances in Convolutional Neural Network Acceleration (2018, background): “91× speed-up on mobile device with less than 1% error rate increase [51].”
Coordinating Filters for Faster Deep Neural Networks (2017, methodology): “Similar evaluations on mobile devices were also reported [14][27].”
A review of convolutional neural network architectures and their optimizations (2022, background): “Wang and Cheng (2016) proposed a CNNs acceleration technique called block term decomposition (BTD) in 2016.”
Towards Accurate Post-training Network Quantization via Bit-Split and Stitching (2020, background): “Thus network compression (Han et al., 2015; Wu et al., 2016; Wang & Cheng, 2016; Romero et al., 2014) has drawn great attention of researchers.”