Research map / Architectures and efficiency / CNN design
Geometric Neural Phrase Pooling: Modeling the Spatial Co-occurrence of Neurons
Lingxi Xie, Qi Tian, John Flynn, Jingdong Wang, Alan Yuille
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add GNPP before pooling layers | 3-layer LeNet · CIFAR10 | error rate (%): 17.07 -> 14.78 | Improved |
| add GNPP before pooling layers 2 and 3 | 3-layer LeNet · CIFAR100 | error rate (%): 44.99 -> 42.04 | Improved |
| add GNPP before pooling layers | 3-layer LeNet · SVHN | error rate (%): 4.63 -> 3.39 | Improved |
| add GNPP before pooling layers | 2-layer LeNet · MNIST | error rate (%): 0.87 -> 0.63 | Improved |
| add GNPP on top of Dropout | 2-layer LeNet · MNIST | error rate (%): 0.72 -> 0.58 | Improved |
| replace GNPP with vanilla Gaussian blur before pooling (ablation) | 3-layer LeNet · CIFAR10 | error rate (%): 17.07 -> 17.05 | No clear effect |
| add GNPP before pooling layers 2 and 3 | BigNet · CIFAR10 | error rate (%): 7.80 -> 7.14 | Improved |
| add GNPP before last pooling layer | WRN · CIFAR100 | error rate (%): 25.52 -> 25.01 | Improved |
| add GNPP before pool-5 | AlexNet · ImageNet top-1 error | error rate (%): 43.19 -> 42.16 | Improved |
| add GNPP before pool-1 or pool-2 | AlexNet · ImageNet | error rate (%) | Hurt |
| insert GNPP before last pooling layer only instead of first (ablation) | 3-layer LeNet · CIFAR10 | error rate (%): 16.67 -> 15.49 | Improved |
| insert GNPP before all three pooling layers instead of layers 2 and 3 (ablation) | 3-layer LeNet · SVHN | error rate (%): 3.43 -> 3.46 | Hurt |
Closest papers by meaning
- Geometric Deep Learning on Graphs and Manifolds Using Mixture Model CNNs (CVPR 2017 · similarity 0.85)
- Learning Visual N-Grams From Web Data (ICCV 2017 · similarity 0.84)
- Revisiting Image-Language Networks for Open-Ended Phrase Detection (TPAMI 2020 · similarity 0.84)
- Global Second-Order Pooling Convolutional Networks (CVPR 2019 · similarity 0.83)
- View N-Gram Network for 3D Object Retrieval (ICCV 2019 · similarity 0.83)
- LINE: LLM-based Iterative Neuron Explanations for Vision Models (NeurIPS 2026 · similarity 0.83)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
