Research map / Architectures and efficiency / RNNs and normalization
Towards Efficient Tensor Decomposition-Based DNN Model Compression With Optimization Framework
Check 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 ADMM-based TT compression instead of direct TT/TR training | LeNet-5 · MNIST | Top-1 accuracy (%): 99.07 -> 99.48 | Mixed |
| Use ADMM-based TT compression instead of the strongest reported TT/TR baseline | ResNet-20 · CIFAR-10 | Top-1 accuracy (%): 90.80 -> 91.47 | Helped |
| Use ADMM-based TT compression instead of the strongest reported TT/TR baseline | ResNet-32 · CIFAR-10 | Top-1 accuracy (%): 91.44 -> 92.87 | Helped |
| Use ADMM-based TT compression instead of the strongest reported TT/TR baseline | ResNet-20 · CIFAR-100 | Top-1 accuracy (%): 66.13 -> 67.36 | Helped |
| Use ADMM-based TT compression instead of the strongest reported TT/TR baseline | ResNet-32 · CIFAR-100 | Top-1 accuracy (%): 68.05 -> 70.31 | Helped |
| Use ADMM-based TT compression instead of standard TT training | ResNet-18 · ImageNet | Top-5 accuracy (%): 85.64 -> 87.47 | Helped |
| Use ADMM-based TT compression instead of TR-LSTM | LSTM · UCF11 | Top-1 accuracy (%): +2.1% | Helped |
| Use ADMM-based TT compression instead of TR-LSTM | LSTM · HMDB51 | Top-1 accuracy (%): +0.29% | Helped |
| Set ADMM penalty parameter rho too small (0.001) (ablation) | ADMM-regularized ResNet-32 training · CIFAR-10 | Low TT-rank characteristics | Hurt |
| Use smaller rho (ablation) | ADMM-regularized ResNet-32 training · CIFAR-10 | Test accuracy | Helped |
| Vary rho across tested settings (ablation) | ADMM-regularized ResNet-32 training · CIFAR-10 | Convergence speed | No clear effect |
Closest papers by meaning
- BATUDE: Budget-Aware Neural Network Compression Based on Tucker Decomposition (AAAI 2022 · similarity 0.91)
- Distribution-Aware Tensor Decomposition for Compression of Convolutional Neural Networks (NeurIPS 2025 · similarity 0.90)
- Exploring Unexplored Tensor Network Decompositions for Convolutional Neural Networks (NeurIPS 2019 · similarity 0.90)
- Wide Compression: Tensor Ring Nets (CVPR 2018 · similarity 0.89)
- "Lossless" Compression of Deep Neural Networks: A High-dimensional Neural Tangent Kernel Approach (NeurIPS 2022 · similarity 0.89)
- ESPACE: Dimensionality Reduction of Activations for Model Compression (NeurIPS 2024 · similarity 0.86)
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.