Research map / Learning theory and algorithms / Matrix and tensor methods
Dropout: Explicit Forms and Capacity Control
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 |
|---|---|---|---|
| Apply dropout at rate 0.1 instead of selecting the best plain-SGD iterate | Matrix factorization, width 30 · MovieLens test | RMSE: 0.7938 -> 0.7805 | Helped |
| Apply dropout at rate 0.3 instead of selecting the best plain-SGD iterate | Matrix factorization, width 70 · MovieLens test | RMSE: 0.7897 -> 0.7763 | Helped |
| Apply dropout at rate 0.3 instead of selecting the best plain-SGD iterate | Matrix factorization, width 110 · MovieLens test | RMSE: 0.7873 -> 0.7742 | Helped |
| Apply dropout at rate 0.4 instead of selecting the best plain-SGD iterate | Matrix factorization, width 150 · MovieLens test | RMSE: 0.7858 -> 0.7722 | Helped |
| Apply dropout at rate 0.4 instead of selecting the best plain-SGD iterate | Matrix factorization, width 190 · MovieLens test | RMSE: 0.7844 -> 0.772 | Helped |
| Increase the dropout rate | Two-layer neural networks · MNIST binary classification, classes 4 and 7 | generalization gap | Helped |
| Increase the dropout rate | Two-layer neural networks · MNIST binary classification, classes 4 and 7 | co-adaptation | Helped |
| Increase the dropout rate | Matrix factorization · MovieLens | training error | Hurt |
| Train with randomized sign symmetrization | Two-layer neural networks · MNIST and Fashion-MNIST | performance | No clear effect |
Closest papers by meaning
- Adaptive Dropout with Rademacher Complexity Regularization (ICLR 2018 · similarity 0.90)
- On Dropout and Nuclear Norm Regularization (ICML 2019 · similarity 0.90)
- On the Regularization Properties of Structured Dropout (CVPR 2020 · similarity 0.90)
- Dropout Training, Data-dependent Regularization, and Generalization Bounds (ICML 2018 · similarity 0.89)
- Demystifying Dropout (ICML 2019 · similarity 0.89)
- Stochastic Modified Equations and Dynamics of Dropout Algorithm (ICLR 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.