mlsanityData to 14 Oct 2026

Neural expressiveness for beyond importance model compression

arXiv 2512.06440 ↗First posted 6 Dec 2025cs.LGcited by 0

Not on the map. Why?

Abstract (arXiv)

Neural Network Pruning has been established as driving force in the exploration of memory and energy efficient solutions with high throughput both during training and at test time. In this paper, we introduce a novel criterion for model compression, named "Expressiveness". Unlike existing pruning methods that rely on the inherent "Importance" of neurons' and filters' weights, ``Expressiveness" emphasizes a neuron's or group of neurons ability to redistribute informational resources effectively, based on the overlap of activations. This characteristic is strongly correlated to a network's initialization state, establishing criterion autonomy from the learning state stateless and thus setting a new fundamental basis for the expansion of compression strategies in regards to the "When to Prune" question. We show that expressiveness is effectively approximated with arbitrary data or limited dataset's representative samples, making ground for the exploration of Data-Agnostic strategies. Our work also facilitates a "hybrid" formulation of expressiveness and importance-based pruning strategies, illustrating their complementary benefits and delivering up to 10x extra gains w.r.t. weight-based approaches in parameter compression ratios, with an average of 1% in performance degradation. We also show that employing expressiveness (independently) for pruning leads to an improvement over top-performing and foundational methods in terms of compression efficiency. Finally, on YOLOv8, we achieve a 46.1% MACs reduction by removing 55.4% of the parameters, with an increase of 3% in the mean Absolute Precision (mAP50-95) for object detection on COCO dataset.

Key facts. No citing papers in the map yet (data through 14 Oct 2026). Builds on 46 of its 61 references that are in the map. First posted on arXiv 6 Dec 2025.

Builds on

46 of 61 references.

References · 46

  1. 2015Deep Residual Learning for Image Recognitioncited by 52,096
  2. 2014Microsoft COCO: Common Objects in Contextcited by 17,608
  3. 2014Very Deep Convolutional Networks for Large-Scale Image Recognitioncited by 15,364
  4. 2014ImageNet Large Scale Visual Recognition Challengecited by 8,498
  5. 2016Densely Connected Convolutional Networkscited by 6,720
  6. 2014Going Deeper with Convolutionscited by 4,241
  7. 2017MobileNets: Efficient Convolutional Neural Networks for Mobile Vision Applicationscited by 3,845
  8. 2015Deep Compression: Compressing Deep Neural Networks with Pruning, Trained Quantization and Huffman Codingcited by 2,329
  9. 2018The Lottery Ticket Hypothesis: Finding Sparse, Trainable Neural Networkscited by 2,262
  10. 2015Learning both Weights and Connections for Efficient Neural Networkscited by 2,031
  11. 2017ShuffleNet: An Extremely Efficient Convolutional Neural Network for Mobile Devicescited by 1,200
  12. 2016Pruning Filters for Efficient ConvNetscited by 1,143
34 more
  1. 2018SNIP: Single-shot Network Pruning based on Connection Sensitivitycited by 812
  2. 2017Learning Efficient Convolutional Networks through Network Slimmingcited by 676
  3. 2018Rethinking the Value of Network Pruningcited by 599
  4. 2020What is the State of Neural Network Pruning?cited by 583
  5. 2019Importance Estimation for Neural Network Pruningcited by 541
  6. 2020Pruning neural networks without any data by iteratively conserving synaptic flowcited by 479
  7. 2020Picking Winning Tickets Before Training by Preserving Gradient Flowcited by 471
  8. 2017ThiNet: A Filter Level Pruning Method for Deep Neural Network Compressioncited by 442
  9. 2020Movement Pruning: Adaptive Sparsity by Fine-Tuningcited by 403
  10. 2020HRank: Filter Pruning using High-Rank Feature Mapcited by 292
  11. 2016Dynamic Network Surgery for Efficient DNNscited by 233
  12. 2017NISP: Pruning Networks using Neuron Importance Score Propagationcited by 228
  13. 2016Network Trimming: A Data-Driven Neuron Pruning Approach towards Efficient Deep Architecturescited by 221
  14. 2023DepGraph: Towards Any Structural Pruningcited by 203
  15. 2018Discrimination-aware Channel Pruning for Deep Neural Networkscited by 171
  16. 2019Towards Optimal Structured CNN Pruning via Generative Adversarial Learningcited by 146
  17. 2016Designing Energy-Efficient Convolutional Neural Networks using Energy-Aware Pruningcited by 134
  18. 2020Layer-adaptive sparsity for the Magnitude-based Pruningcited by 126
  19. 2018Faster gaze prediction with dense networks and Fisher pruningcited by 94
  20. 2017Shift: A Zero FLOP, Zero Parameter Alternative to Spatial Convolutionscited by 93
  21. 2020Channel Pruning via Automatic Structure Searchcited by 85
  22. 2020Operation-Aware Soft Channel Pruning using Differentiable Maskscited by 74
  23. 2019Global Sparse Momentum SGD for Pruning Very Deep Neural Networkscited by 70
  24. 2020Methods for Pruning Deep Neural Networkscited by 49
  25. 2021Recent Advances on Neural Network Pruning at Initializationcited by 41
  26. 2018Exploiting Kernel Sparsity and Entropy for Interpretable CNN Compressioncited by 40
  27. 2021When to Prune? A Policy towards Early Structural Pruningcited by 38
  28. 2020Layer Pruning via Fusible Residual Convolutional Block for Deep Neural Networkscited by 18
  29. 2023Comprehensive Multimodal Segmentation in Medical Imaging: Combining YOLOv8 with SAM and HQ-SAM Modelscited by 18
  30. 2023Efficient Joint Optimization of Layer-Adaptive Weight Pruning in Deep Neural Networkscited by 15
  31. 2023Real-time Multi-Class Helmet Violation Detection Using Few-Shot Data Sampling Technique and YOLOv8cited by 13
  32. 2023Differentiable Transportation Pruningcited by 10
  33. 2019Online Filter Clustering and Pruning for Efficient Convnetscited by 6
  34. 2025Rip Current Segmentation: A Novel Benchmark and YOLOv8 Baseline Resultscited by 4

Built on by

Not cited yet.