Research map / Architectures and efficiency / RNNs and normalization
Efficient Neural Network Compression
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 whole-network rank selection with ENC-Inf instead of layer-by-layer compression | VGG-16 · ImageNet, FLOPs 20% | Top-5 ΔAcc.: 0.3% vs 0.0% | Helped |
| Use whole-network rank selection with ENC-Inf instead of layer-by-layer compression | VGG-16 · ImageNet, FLOPs 25% | Top-1 ΔAcc. w/o FT: 13.0 -> 10.7 | Helped |
| Use whole-network rank selection with ENC-Inf instead of layer-by-layer compression | VGG-16 · ImageNet, FLOPs 25% | Top-5 ΔAcc. w/o FT: 9.1 -> 7.3 | Helped |
| Use ENC-Inf with validation-set inference rather than ENC-Map | VGG-16 · ImageNet, high compression | accuracy | Helped |
| Use ENC-Map rather than ENC-Inf or ENC-Model | VGG-16 · ImageNet, low compression | accuracy | No clear effect |
| Use ENC-Inf instead of the Uniform rank-reduction baseline | ResNet-56 · Cifar-10, FLOPs 50%, w/o fine-tuning | Top-1 ΔAcc.: 12.7 -> 2.9 | Helped |
| Use measurement-based layer-wise accuracy metrics rather than PCA energy for shallow-network rank choice | AlexNet · ImageNet | accuracy | Helped |
| Use PCA energy-based rather than measurement-based layer-wise accuracy metrics | VGG-16 and ResNet-56 · ImageNet and Cifar-10 | accuracy | No clear effect |
| Fine-tune the compressed ResNet-56 | ENC-Inf · Cifar-10, FLOPs 50% | Top-1 ΔAcc.: 2.9 -> 0.1 | Helped |
| Fine-tune the compressed VGG-16 | ENC-Model · ImageNet, FLOPs 25% | Top-1 ΔAcc.: 11.3 -> -0.7 | Helped |
| Allow one fine-tuning epoch instead of no fine-tuning | ENC-Model · VGG-16, lossless compression | FLOPs: 57% -> 33% | Helped |
Closest papers by meaning
- Compressing Neural Networks: Towards Determining the Optimal Layer-wise Decomposition (NeurIPS 2021 · similarity 0.89)
- Low-Rank Compression of Neural Nets: Learning the Rank of Each Layer (CVPR 2020 · similarity 0.89)
- "Lossless" Compression of Deep Neural Networks: A High-dimensional Neural Tangent Kernel Approach (NeurIPS 2022 · similarity 0.89)
- HALOC: Hardware-Aware Automatic Low-Rank Compression for Compact Neural Networks (AAAI 2023 · similarity 0.88)
- Data-Dependent Coresets for Compressing Neural Networks with Applications to Generalization Bounds (ICLR 2019 · similarity 0.87)
- Distribution-Aware Tensor Decomposition for Compression of Convolutional Neural Networks (NeurIPS 2025 · 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.