Research map / Architectures and efficiency / RNNs and normalization
Curriculum Dropout
Pietro Morerio, Jacopo Cavazza, Riccardo Volpi, Rene Vidal, Vittorio Murino
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (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 Curriculum Dropout instead of standard Dropout | Dropout · Double MNIST (CNN-2, n fixed) | accuracy gain over unregularized: +1.42 -> +2.35 | Improved |
| use Curriculum Dropout instead of standard Dropout | Dropout · CIFAR-10 (CNN-1) | accuracy gain over unregularized: +0.22 -> +0.62 | Improved |
| use Curriculum Dropout instead of standard Dropout | Dropout · CIFAR-100 (CNN-1) | accuracy gain over unregularized: +1.01 -> +1.66 | Improved |
| use Curriculum Dropout instead of standard Dropout | Dropout · Caltech-256 (CNN-2) | accuracy gain over unregularized: +2.36 -> +3.23 | Improved |
| use Curriculum Dropout instead of standard Dropout | Dropout · Caltech-101 (CNN-2) | accuracy gain over unregularized: +4.21 -> +4.72 | Improved |
| use Curriculum Dropout instead of standard Dropout | Dropout · SVHN (CNN-2, n fixed) | accuracy gain over unregularized: +2.35 -> +2.65 | Improved |
| use Curriculum Dropout instead of standard Dropout | Dropout · MNIST (MLP) | accuracy gain over unregularized: +0.38 -> +0.36 | No clear effect |
| use Curriculum Dropout instead of standard Dropout | Dropout · MNIST (CNN-1) | accuracy gain over unregularized: +0.15 -> +0.18 | Improved |
| use decreasing retain probability (curriculum) instead of linearly increasing (anti-curriculum) (ablation) | Anti-Curriculum Dropout · CIFAR-10 (CNN-1) | accuracy gain over unregularized: -0.68 -> +0.62 | Improved |
| use decreasing retain probability (curriculum) instead of linearly increasing (anti-curriculum) (ablation) | Anti-Curriculum Dropout · Caltech-256 (CNN-2) | accuracy gain over unregularized: -0.22 -> +3.23 | Improved |
| use Curriculum Dropout instead of anti-curriculum (ablation) | Anti-Curriculum Dropout · Double MNIST (CNN-2, n fixed) | accuracy gain over unregularized: +0.73 -> +2.35 | Improved |
| use smooth exponential schedule instead of a Switch-Curriculum step function (ablation) | Curriculum Dropout · Double MNIST | test accuracy | Improved |
Closest papers by meaning
- Implicit Regularization of Dropout (TPAMI 2024 · similarity 0.91)
- Dropout Reduces Underfitting (ICML 2023 · similarity 0.90)
- Advanced Dropout: A Model-free Methodology for Bayesian Dropout Optimization (TPAMI 2021 · similarity 0.90)
- On the Regularization Properties of Structured Dropout (CVPR 2020 · similarity 0.89)
- Stochastic Modified Equations and Dynamics of Dropout Algorithm (ICLR 2024 · similarity 0.89)
- Activation by Interval-wise Dropout: A Simple Way to Prevent Neural Networks from Plasticity Loss (ICML 2025 · similarity 0.87)
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.
