Research map / Architectures and efficiency / CNN design
PR Product: A Substitute for Inner Product in Neural Networks
Zhennan Wang, Wenbin Zou, Chen Xu
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 |
|---|---|---|---|
| replace inner product with PR Product | ResNet110 · CIFAR10 | error rate: 6.23 -> 5.97 | Improved |
| replace inner product with PR Product | PreResNet110 · CIFAR10 | error rate: 5.99 -> 5.64 | Improved |
| replace inner product with PR Product | WRN-28-10 · CIFAR10 | error rate: 4.34 -> 4.03 | Improved |
| replace inner product with PR Product | DenseNet-BC-100-12 · CIFAR10 | error rate: 4.63 -> 4.46 | Improved |
| replace inner product with PR Product | WRN-28-10 · CIFAR100 | error rate: 19.50 -> 19.57 | No clear effect |
| replace inner product with PR Product | ResNet110, PreResNet110, DenseNet-BC-100-12 · CIFAR100 | error rate: 28.08 -> 27.88 | Improved |
| use PR Product instead of standard inner product | captioning model · MS COCO Karpathy test (cross-entropy) | CIDEr: 116.0 -> 116.1 | Improved |
| use PR Product instead of standard inner product | captioning model · MS COCO Karpathy test (CIDEr optimization) | CIDEr: 127.2 -> 128.3 | Improved |
| use only the R Product (rejection) instead of the P Product (ablation) | captioning model · MS COCO Karpathy test (cross-entropy) | CIDEr: 116.0 -> 113.3 | Hurt |
| use only the R Product (rejection) instead of the P Product (ablation) | captioning model · MS COCO Karpathy test (CIDEr optimization) | CIDEr: 127.2 -> 125.5 | Hurt |
| combine projection and rejection (PR) instead of rejection alone (ablation) | captioning model · MS COCO Karpathy test (cross-entropy) | BLEU-4: 36.0 -> 37.0 | Improved |
| combine projection and rejection (PR) instead of rejection alone (ablation) | captioning model · MS COCO Karpathy test (CIDEr optimization) | BLEU-4: 37.6 -> 38.7 | Improved |
Closest papers by meaning
- Orthogonal Projection Loss (ICCV 2021 · similarity 0.83)
- Attentive Tensor Product Learning (AAAI 2019 · similarity 0.82)
- Decoupled Networks (CVPR 2018 · similarity 0.82)
- Prospect Pruning: Finding Trainable Weights at Initialization using Meta-Gradients (ICLR 2022 · similarity 0.82)
- Entropy Induced Pruning Framework for Convolutional Neural Networks (AAAI 2024 · similarity 0.81)
- Attention-based Iterative Decomposition for Tensor Product Representation (ICLR 2024 · similarity 0.81)
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.
