Learning to Prune Filters in Convolutional Neural Networks

By Qiangui Huang, Shaohua Kevin Zhou, Suya You, Ulrich Neumann. 2018

186 citations recorded by the source corpus.

Many state-of-the-art computer vision algorithms use large scale convolutional neural networks (CNNs) as basic building blocks. These CNNs are known for their huge number of parameters, high redundancy in weights, and tremendous computing resource consumptions. This paper presents a learning algorithm to simplify and speed up these CNNs. Specifically, we introduce a “try-and-learn” algorithm to train pruning agents that remove unnecessary CNN filters in a data-driven way. With the help of a novel reward function, our agents removes a significant number of filters in CNNs while maintaining performance at a desired level. Moreover, this method provides an easy control of the tradeoff between network performance and its scale. Performance of our algorithm is validated with comprehensive pruning experiments on several popular CNNs for visual recognition and semantic segmentation tasks.

Subfields: Computer Vision and Pattern Recognition, Efficiency & Compression

How researchers see this paper

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

In Computer Vision and Pattern Recognition

In Efficiency & Compression

Citations in the corpus

AutoSlim: Towards One-Shot Architecture Search for Channel Numbers (2019, methodology): “, network slimming) methods [1, 30, 39, 40, 41] aim at reducing effective channels of a large neural network to speedup its inference.”
Network Slimming by Slimmable Networks: Towards One-Shot Architecture Search for Channel Numbers (2019, methodology): “, network slimming) methods [1, 30, 39, 40, 41] aim at reducing effective channels of a large neural network to speedup its inference.”
YOLOv4-5D: An Effective and Efficient Object Detector for Autonomous Driving (2021, background): “By regularizing the scaling factor L1 and continuously pushing it toward 0, the convolution kernel corresponding to the scaling factor close to 0 will slightly contribute to the network inference [42].”
Cascaded Projection: End-To-End Network Compression and Acceleration (2019, methodology): “Classification accuracy drop on CIFAR10, relative to baseline, of compression methods (CaP, PCAS [53], PFEC [35] and LPF [24]) for a range of compression levels on ResNet18 (Top) and ResNet50 (Bottom).”
Filter Pruning via Geometric Median for Deep Convolutional Neural Networks Acceleration (2018, methodology): “Some filter pruning approaches [23, 25, 16, 27, 7, 33, 39, 37, 46, 14, 18, 22] need to utilize training data to determine the pruned filters.”
Meta Filter Pruning to Accelerate Deep Convolutional Neural Networks (2019, methodology): “[19, 20, 15, 21, 6, 29, 33, 31, 37, 16, 12] utilize the training data to determine the pruned filters.”
Targeted Dropout (2018, background): “There has been a great deal of work on developing strategies to sparsify neural networks [10, 7, 5, 2, 13, 3, 12, 9, 18].”
Efficient Acceleration of Deep Learning Inference on Resource-Constrained Edge Devices: A Review (2023, methodology): “A structured pruning technique presented in [100] for CNN removes certain filters to increase the efficiency of the network without significantly affecting the accuracy.”
Dynamic Channel Pruning: Feature Boosting and Suppression (2018, methodology): “(2016); Alvarez & Salzmann (2016); Li et al. (2017); He et al.”