mlsanityData to 7 Oct 2026

Designing Semi-Structured Pruning of Graph Convolutional Networks for Skeleton-based Recognition

arXiv 2412.11813 ↗First posted 16 Dec 2024cs.CVcited by 0

Not on the map. Why?

Abstract (arXiv)

Deep neural networks (DNNs) are nowadays witnessing a major success in solving many pattern recognition tasks including skeleton-based classification. The deployment of DNNs on edge-devices, endowed with limited time and memory resources, requires designing lightweight and efficient variants of these networks. Pruning is one of the lightweight network design techniques that operate by removing unnecessary network parts, in a structured or an unstructured manner, including individual weights, neurons or even entire channels. Nonetheless, structured and unstructured pruning methods, when applied separately, may either be inefficient or ineffective. In this paper, we devise a novel semi-structured method that discards the downsides of structured and unstructured pruning while gathering their upsides to some extent. The proposed solution is based on a differentiable cascaded parametrization which combines (i) a band-stop mechanism that prunes weights depending on their magnitudes, (ii) a weight-sharing parametrization that prunes connections either individually or group-wise, and (iii) a gating mechanism which arbitrates between different group-wise and entry-wise pruning. All these cascaded parametrizations are built upon a common latent tensor which is trained end-to-end by minimizing a classification loss and a surrogate tensor rank regularizer. Extensive experiments, conducted on the challenging tasks of action and hand-gesture recognition, show the clear advantage of our proposed semi-structured pruning approach against both structured and unstructured pruning, when taken separately, as well as the related work.

Key facts. No citing papers in the map yet (data through 7 Oct 2026). Builds on 32 of its 97 references that are in the map. First posted on arXiv 16 Dec 2024.

Builds on

32 of 97 references.

References · 32

  1. 2016Semi-Supervised Classification with Graph Convolutional Networkscited by 11,282
  2. 2015Deep Compression: Compressing Deep Neural Networks with Pruning, Trained Quantization and Huffman Codingcited by 2,291
  3. 2015Learning both Weights and Connections for Efficient Neural Networkscited by 1,998
  4. 2019Searching for MobileNetV3cited by 1,984
  5. 2018Spatial Temporal Graph Convolutional Networks for Skeleton-Based Action Recognitioncited by 1,199
  6. 2016Pruning Filters for Efficient ConvNetscited by 1,129
  7. 2017Learning Efficient Convolutional Networks through Network Slimmingcited by 673
  8. 2017Learning Sparse Neural Networks through L0 Regularizationcited by 571
  9. 2019Improved Knowledge Distillation via Teacher Assistantcited by 567
  10. 2016Learning Structured Sparsity in Deep Neural Networkscited by 538
  11. 2016Convolutional Two-Stream Network Fusion for Video Action Recognitioncited by 384
  12. 2017First-Person Hand Action Benchmark with RGB-D Videos and 3D Hand Pose Annotationscited by 229
20 more
  1. 2019Understanding Attention and Generalization in Graph Neural Networkscited by 169
  2. 2016A Riemannian Network for SPD Matrix Learningcited by 162
  3. 2018Adaptive Graph Convolutional Neural Networkscited by 155
  4. 2016Spatio-Temporal LSTM with Trust Gates for 3D Human Action Recognitioncited by 142
  5. 2016An End-to-End Spatio-Temporal Attention Model for Human Action Recognition from Skeleton Datacited by 126
  6. 2017View Adaptive Recurrent Neural Networks for High Performance Human Action Recognition from Skeleton Datacited by 105
  7. 2017MorphNet: Fast & Simple Resource-Constrained Structure Learning of Deep Networkscited by 92
  8. 2016Co-occurrence Feature Learning for Skeleton based Action Recognition using Regularized Deep LSTM Networkscited by 63
  9. 2017Skeleton-Based Human Action Recognition with Global Context-Aware Attention LSTM Networkscited by 62
  10. 2016Building Deep Networks on Grassmann Manifoldscited by 58
  11. 2017RGB-D-based Human Motion Recognition with Deep Learning: A Surveycited by 37
  12. 2022Pruning-as-Search: Efficient Neural Architecture Search via Channel Pruning and Structural Reparameterizationcited by 31
  13. 2018Structured Pruning of Neural Networks with Budget-Aware Regularizationcited by 26
  14. 2018A Novel Geometric Framework on Gram Matrix Trajectories for Human Behavior Understandingcited by 21
  15. 2021Learning Connectivity with Graph Convolutional Networks for Skeleton-based Action Recognitioncited by 20
  16. 2016Transition Forests: Learning Discriminative Temporal Transitions for Action Recognition and Detectioncited by 18
  17. 2018Spatio-Temporal Graph Convolution for Skeleton Based Action Recognitioncited by 15
  18. 2018DeepGRU: Deep Gesture Recognition Utilitycited by 14
  19. 2016DropNeuron: Simplifying the Structure of Deep Neural Networkscited by 11
  20. 2018Entropy-Constrained Training of Deep Neural Networkscited by 8

Built on by

Not cited yet.