Research map / Optimization / Optimizers
Improved Dropout for Shallow and Deep Learning
Zhe Li, Boqing Gong, Tianbao Yang
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 |
|---|---|---|---|
| replace standard Bernoulli dropout with evolutional dropout (mini-batch second-order sampling probabilities) | CNN on CIFAR-100 · CIFAR-100 | testing performance: over 10% relative improvement | Improved |
| replace standard dropout with evolutional dropout | CNN on CIFAR-100 · CIFAR-100 | convergence speed: over 50% | Improved |
| replace standard dropout with data-dependent dropout | logistic regression · real-sim, news20, RCV1 | training/testing error convergence | Improved |
| replace standard dropout with evolutional dropout | CNNs on MNIST, SVHN, CIFAR-10, CIFAR-100 · MNIST, SVHN, CIFAR-10, CIFAR-100 | convergence speed | Improved |
| run SVHN training beyond 12000 iterations | evolutional vs standard dropout · SVHN | error curves | No clear effect |
| evolutional dropout instead of BN plus standard dropout | Caffe CIFAR-10 network · CIFAR-10 | testing accuracy | No clear effect |
| use sigmoid activation with BN instead of ReLU (ablation) | BN network · CIFAR-10 | testing accuracy | Hurt |
Closest papers by meaning
- Advanced Dropout: A Model-free Methodology for Bayesian Dropout Optimization (TPAMI 2021 · similarity 0.90)
- Demystifying Dropout (ICML 2019 · similarity 0.90)
- Dropout with Expectation-linear Regularization (ICLR 2017 · similarity 0.89)
- Dropout as a Bayesian Approximation: Representing Model Uncertainty in Deep Learning (ICML 2016 · similarity 0.89)
- Excitation Dropout: Encouraging Plasticity in Deep Neural Networks (IJCV 2021 · similarity 0.88)
- Stochastic Modified Equations and Dynamics of Dropout Algorithm (ICLR 2024 · 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.
