Research map / Architectures and efficiency / CNN design
Wide Compression: Tensor Ring Nets
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| Use tensor-ring factorization instead of matrix factorization | LeNet-300-100 · MNIST | error rate (%): 3.91 -> 3.75 | Mixed |
| Increase tensor-ring rank from r=3 to r=5 (ablation) | TRN LeNet-300-100 · MNIST | error rate (%): 8.53 -> 3.75 | Helped |
| Increase tensor-ring rank from r=5 to r=15 (ablation) | TRN LeNet-300-100 · MNIST | error rate (%): 3.75 -> 2.64 | Helped |
| Increase tensor-ring rank from r=3 to r=20 (ablation) | TRN LeNet-5 · MNIST | error rate (%): 2.24 -> 0.69 | Helped |
| Use TRN compression at r=20 rather than the uncompressed LeNet-5 | LeNet-5 · MNIST | error rate (%): 0.79 -> 0.69 | Helped |
| Use TRN compression at r=3 rather than the uncompressed LeNet-5 | LeNet-5 · MNIST | error rate (%): 0.79 -> 2.24 | Hurt |
| Use tensor-ring rather than Tucker factorization at comparable 5× compression | ResNet-32L · CIFAR10 | error rate (%): 12.3 -> 9.4 | Helped |
| Increase tensor-ring rank from r=2 to r=10 (ablation) | TRN ResNet-32L · CIFAR10 | error rate (%): 22.2 -> 9.4 | Helped |
| Increase tensor-ring rank from r=2 to r=10 (ablation) | TRN WideResNet-28L · CIFAR10 | error rate (%): 16.3 -> 7.3 | Helped |
| Compress WideResNet using TRN at r=10 | WideResNet-28L · CIFAR10 | error rate (%): 5.0 -> 7.3 | Hurt |
| Use tensor-ring rather than Tucker factorization at comparable 5× compression | ResNet-32L · CIFAR100 | error rate (%): 42.2 -> 33.3 | Helped |
| Increase tensor-ring rank from r=2 to r=6 (ablation) | TRN ResNet-32L · CIFAR100 | error rate (%): 51.3 -> 36.6 | Helped |
Closest papers by meaning
- "Lossless" Compression of Deep Neural Networks: A High-dimensional Neural Tangent Kernel Approach (NeurIPS 2022 · similarity 0.89)
- Towards Efficient Tensor Decomposition-Based DNN Model Compression With Optimization Framework (CVPR 2021 · similarity 0.89)
- Geometry-aware training of factorized layers in tensor Tucker format (NeurIPS 2024 · similarity 0.86)
- Distribution-Aware Tensor Decomposition for Compression of Convolutional Neural Networks (NeurIPS 2025 · similarity 0.86)
- Layer Compression of Deep Networks with Straight Flows (AAAI 2024 · similarity 0.85)
- Convolutions and More as Einsum: A Tensor Network Perspective with Advances for Second-Order Methods (NeurIPS 2024 · similarity 0.85)
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.