Research map / Architectures and efficiency / Pruning and efficient tuning
Validating the Lottery Ticket Hypothesis with Inertial Manifold Theory
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 IMC pruning instead of LTH iterative magnitude pruning (reinitialization) at the same sparsity | ResNet-20 · CIFAR-10 | accuracy: 92.9 -> 94.5 | Helped |
| use IMC one-time pruning instead of LTH+Reinitialization | ResNet-20 · CIFAR-10 | runtime (s): 24,901 -> 3,215 | Helped |
| use IMC instead of SNIP/SynFlow/GraSP (pruning before training) in runtime | ResNet-20 · CIFAR-10 | runtime (s): 2,664 -> 3,215 | Hurt |
| swap subnetwork training strategy among reinitialization, rewinding, fine-tuning (ablation) | IMC · CIFAR-10 / ImageNet | accuracy: 94.4-95.1 (CIFAR-10); 59.2-60.7 (ImageNet) | No clear effect |
| vary number of training epochs of original network used to estimate W* (ablation) | IMC · CIFAR-10, ImageNet, CIFAR-100 | accuracy: within <1% fluctuation (e.g. 94.18-94.57 for IMC-Reinitialization on CIFAR-10) | No clear effect |
| increase sparsity beyond the IMC-found level (ablation) | IMC · CIFAR-10, ImageNet, CIFAR-100 | accuracy | Hurt |
| IMC pruning vs dense baseline | ResNet-32 · ImageNet | accuracy: 60.0 -> 60.9 | Helped |
| IMC pruning vs dense baseline | ResNet-20 · CIFAR-10 | accuracy: 94.6 -> 94.6 | No clear effect |
Closest papers by meaning
- The Lottery Ticket Hypothesis: Finding Sparse, Trainable Neural Networks (ICLR 2019 · similarity 0.93)
- Why Lottery Ticket Wins? A Theoretical Perspective of Sample Complexity on Sparse Neural Networks (NeurIPS 2021 · similarity 0.92)
- Polynomially Over-Parameterized Convolutional Neural Networks Contain Structured Strong Winning Lottery Tickets (NeurIPS 2023 · similarity 0.91)
- Find A Winning Sign: Sign Is All We Need to Win the Lottery (ICLR 2025 · similarity 0.90)
- No Free Prune: Information-Theoretic Barriers to Pruning at Initialization (ICML 2024 · similarity 0.90)
- On the Sparsity of the Strong Lottery Ticket Hypothesis (NeurIPS 2024 · similarity 0.90)
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.