Toward Compact ConvNets via Structure-Sparsity Regularized Filter Pruning

• Published in: IEEE transaction on neural networks and learning systems Vol. 31; no. 2; pp. 574 - 588
• Main Authors: Lin, Shaohui, Ji, Rongrong, Li, Yuchao, Deng, Cheng, Li, Xuelong
• Format: Journal Article
• Language: English
• Published: United States IEEE 01.02.2020; The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
• Subjects: Adaptation models; Artificial neural networks; CNN acceleration; CNN compression; Compression; Computation; Computer vision; Convergence; Convolutional neural networks; Convolutional neural networks (CNNs); Electronic devices; Embedded systems; Feature maps; Filters; Image coding; Image reconstruction; Lagrange multiplier; Libraries; Machine learning; Mobile communication systems; Neural networks; Object detection; Object recognition; Objective function; Optimization; Pruning; Regularization; Sparsity; structured sparsity; Transfer learning
• ISSN: 2162-237X; 2162-2388; 2162-2388
• DOI: 10.1109/TNNLS.2019.2906563

The success of convolutional neural networks (CNNs) in computer vision applications has been accompanied by a significant increase of computation and memory costs, which prohibits their usage on resource-limited environments, such as mobile systems or embedded devices. To this end, the research of CNN compression has recently become emerging. In this paper, we propose a novel filter pruning scheme, termed structured sparsity regularization (SSR), to simultaneously speed up the computation and reduce the memory overhead of CNNs, which can be well supported by various off-the-shelf deep learning libraries. Concretely, the proposed scheme incorporates two different regularizers of structured sparsity into the original objective function of filter pruning, which fully coordinates the global output and local pruning operations to adaptively prune filters. We further propose an alternative updating with Lagrange multipliers (AULM) scheme to efficiently solve its optimization. AULM follows the principle of alternating direction method of multipliers (ADMM) and alternates between promoting the structured sparsity of CNNs and optimizing the recognition loss, which leads to a very efficient solver (<inline-formula> <tex-math notation="LaTeX">2.5\times </tex-math></inline-formula> to the most recent work that directly solves the group sparsity-based regularization). Moreover, by imposing the structured sparsity, the online inference is extremely memory-light since the number of filters and the output feature maps are simultaneously reduced. The proposed scheme has been deployed to a variety of state-of-the-art CNN structures, including LeNet, AlexNet, VGGNet, ResNet, and GoogLeNet, over different data sets. Quantitative results demonstrate that the proposed scheme achieves superior performance over the state-of-the-art methods. We further demonstrate the proposed compression scheme for the task of transfer learning, including domain adaptation and object detection, which also show exciting performance gains over the state-of-the-art filter pruning methods.