Research map / Architectures and efficiency / Pruning and efficient tuning
Lottery Ticket Preserves Weight Correlation: Is It Desirable or Not?
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 learning rate 0.1 instead of 0.01 for pretraining | ResNet-20 · CIFAR-10 | pretraining accuracy: 89.62% -> 91.7% | Helped |
| Use learning rate 0.1 instead of 0.01 for pretraining | MobileNet-V2 · CIFAR-10 | pretraining accuracy: 92.20% vs. 93.86% | Helped |
| Use learning rate 0.1 instead of 0.01 for pretraining | MobileNet-V2 · CIFAR-100 | pretraining accuracy: 73.10% vs. 74.76% | Helped |
| Use learning rate 0.1 instead of 0.01 for pretraining | ResNet-20 · CIFAR-100 | pretraining accuracy: 63.10% vs. 67.15% | Helped |
| Use learning rate 0.1 instead of 0.01 for pretraining | VGG-11 · CIFAR-100 | pretraining accuracy: 67.74% vs. 69.83% | Helped |
| Train the masked original initialization instead of a random reinitialization | ResNet-20 · CIFAR-10, learning rate 0.01 | accuracy | Helped |
| Train the masked original initialization instead of a random reinitialization | ResNet-20 · CIFAR-10, learning rate 0.1 | accuracy | No clear effect |
| Apply the pruning mask to pretrained weights and fine-tune rather than train from the initial ticket | ResNet-20 · CIFAR-10, learning rate 0.1 | accuracy | Helped |
| Use ADMM-based pruning instead of iterative or one-shot pruning to generate the mask (ablation) | Pruning & fine-tuning and sparse training · CIFAR-100 | accuracy | Helped |
| Use one-shot pruning instead of iterative or ADMM-based pruning to generate the mask (ablation) | Pruning & fine-tuning and sparse training · CIFAR-100 | accuracy | Hurt |
| Initialize sparse training from pretrained weights rather than from weights rewound after 10 epochs (ablation) | MobileNet-V2, ResNet-20, and VGG-11 · CIFAR-100, ADMM-based pruning | accuracy | Helped |
| Apply the mask to the pretrained weights rather than to the initialization | ResNet-20 · CIFAR-100, ADMM pruning, sparsity ratio 0.39 | accuracy: 74.76% -> 76.04% | Helped |
Closest papers by meaning
- Why Lottery Ticket Wins? A Theoretical Perspective of Sample Complexity on Sparse Neural Networks (NeurIPS 2021 · similarity 0.92)
- Dual Lottery Ticket Hypothesis (ICLR 2022 · similarity 0.92)
- The Lottery Ticket Hypothesis: Finding Sparse, Trainable Neural Networks (ICLR 2019 · similarity 0.92)
- Plant 'n' Seek: Can You Find the Winning Ticket? (ICLR 2022 · similarity 0.91)
- Find A Winning Sign: Sign Is All We Need to Win the Lottery (ICLR 2025 · similarity 0.90)
- Weisfeiler and Leman Go Gambling: Why Expressive Lottery Tickets Win (ICML 2025 · similarity 0.89)
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.