mlsanityData to 7 Oct 2026

Learnable Heterogeneous Convolution: Learning both topology and strength

arXiv 2301.05440 ↗First posted 13 Jan 2023cs.CVcited by 3

Not on the map. Why?

Abstract (arXiv)

Existing convolution techniques in artificial neural networks suffer from huge computation complexity, while the biological neural network works in a much more powerful yet efficient way. Inspired by the biological plasticity of dendritic topology and synaptic strength, our method, Learnable Heterogeneous Convolution, realizes joint learning of kernel shape and weights, which unifies existing handcrafted convolution techniques in a data-driven way. A model based on our method can converge with structural sparse weights and then be accelerated by devices of high parallelism. In the experiments, our method either reduces VGG16/19 and ResNet34/50 computation by nearly 5x on CIFAR10 and 2x on ImageNet without harming the performance, where the weights are compressed by 10x and 4x respectively; or improves the accuracy by up to 1.0% on CIFAR10 and 0.5% on ImageNet with slightly higher efficiency. The code will be available on www.github.com/Genera1Z/LearnableHeterogeneousConvolution.

Key facts. Cited by 3 papers in the map (data through 7 Oct 2026). Builds on 34 of its 49 references that are in the map. First posted on arXiv 13 Jan 2023.

Builds on

34 of 49 references.

References · 34

  1. 2015Deep Residual Learning for Image Recognitioncited by 51,809
  2. 2014Very Deep Convolutional Networks for Large-Scale Image Recognitioncited by 15,267
  3. 2015Rethinking the Inception Architecture for Computer Visioncited by 5,451
  4. 2018MobileNetV2: Inverted Residuals and Linear Bottleneckscited by 4,773
  5. 2017Squeeze-and-Excitation Networkscited by 4,763
  6. 2014Going Deeper with Convolutionscited by 4,231
  7. 2017MobileNets: Efficient Convolutional Neural Networks for Mobile Vision Applicationscited by 3,827
  8. 2019Searching for MobileNetV3cited by 1,984
  9. 2016Inception-v4, Inception-ResNet and the Impact of Residual Connections on Learningcited by 1,509
  10. 2017ShuffleNet: An Extremely Efficient Convolutional Neural Network for Mobile Devicescited by 1,194
  11. 2018ShuffleNet V2: Practical Guidelines for Efficient CNN Architecture Designcited by 1,132
  12. 2017Learning Sparse Neural Networks through L0 Regularizationcited by 571
22 more
  1. 2019Importance Estimation for Neural Network Pruningcited by 531
  2. 2018Filter Pruning via Geometric Median for Deep Convolutional Neural Networks Accelerationcited by 365
  3. 2019Differentiable Soft Quantization: Bridging Full-Precision and Low-Bit Neural Networkscited by 234
  4. 2019Progressive Differentiable Architecture Search: Bridging the Depth Gap between Search and Evaluationcited by 230
  5. 2019MetaPruning: Meta Learning for Automatic Neural Network Channel Pruningcited by 226
  6. 2017Deep Rewiring: Training very sparse deep networkscited by 136
  7. 2019Orthogonal Convolutional Neural Networkscited by 122
  8. 2017CondenseNet: An Efficient DenseNet using Learned Group Convolutionscited by 121
  9. 2019Dynamic Convolutions: Exploiting Spatial Sparsity for Faster Inferencecited by 91
  10. 2018Shift-Net: Image Inpainting via Deep Feature Rearrangementcited by 87
  11. 2016Understanding and Improving Convolutional Neural Networks via Concatenated Rectified Linear Unitscited by 86
  12. 2019Knowledge Distillation via Route Constrained Optimizationcited by 73
  13. 2014Rigid-Motion Scattering for Texture Classificationcited by 52
  14. 2018Accelerate CNN via Recursive Bayesian Pruningcited by 27
  15. 2019Side Window Filteringcited by 27
  16. 2019HetConv: Heterogeneous Kernel-Based Convolutions for Deep CNNscited by 16
  17. 2018IGCV3: Interleaved Low-Rank Group Convolutions for Efficient Deep Neural Networkscited by 14
  18. 2018IGCV2: Interleaved Structured Sparse Convolutional Neural Networkscited by 12
  19. 2018Extreme Network Compression via Filter Group Approximationcited by 12
  20. 2018BinarEye: An Always-On Energy-Accuracy-Scalable Binary CNN Processor With All Memory On Chip in 28nm CMOScited by 10
  21. 2020ChewBaccaNN: A Flexible 223 TOPS/W BNN Acceleratorcited by 7
  22. 2019Seesaw-Net: Convolution Neural Network With Uneven Group Convolutioncited by 1

Built on by

Cited by year

Citing papers per year: 2024: 3.

3 papers.

  1. 2024Multi-Scale Fusion for Object Representationcited by 6
  2. 2024Grouped Discrete Representation for Object-Centric Learningcited by 5
  3. 2024Organized Grouped Discrete Representation for Object-Centric Learningcited by 1