Research map / Probabilistic and scientific modeling / Uncertainty and Gaussian processes
Contextual Dropout: An Efficient Sample-Dependent Dropout Module
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 sample-dependent Bernoulli dropout probabilities | MLP · Noisy MNIST | accuracy: 86.55 -> 87.43 | Helped |
| use sample-dependent Bernoulli dropout probabilities | MLP · Noisy MNIST | PAvPU: 87.13 -> 87.81 | Helped |
| use sample-dependent Bernoulli dropout probabilities | MLP · Noisy MNIST | log likelihood: -1.68 -> -1.41 | Helped |
| replace stochastic dropout masks with direct gating activation scaling (ablation) | Contextual dropout on MLP · Noisy MNIST | accuracy: 86.20 vs 87.43 | Hurt |
| add regular dropout to gating activations (ablation) | Contextual gating on MLP · Noisy MNIST | accuracy: 86.20 -> 86.70 | Helped |
| replace fine-tuning with training from scratch (ablation) | Contextual dropout on ResNet-18 · ImageNet | Top-1 Acc: 70.03 -> 70.29 | Helped |
| use contextual dropout instead of Gaussian dropout | ResNet-18 · ImageNet | Top-1 Acc: 69.46 -> 70.03 | Helped |
| use contextual dropout instead of regular dropout | WRN · CIFAR-100 original | accuracy: 79.19 -> 80.93 | Helped |
| use contextual dropout instead of regular dropout | MCAN · VQA-v2 noisy | Accuracy: 61.47 -> 63.82 | Helped |
Closest papers by meaning
- Dropout with Expectation-linear Regularization (ICLR 2017 · similarity 0.90)
- Dropout as a Bayesian Approximation: Representing Model Uncertainty in Deep Learning (ICML 2016 · similarity 0.89)
- Dropout Reduces Underfitting (ICML 2023 · similarity 0.89)
- Variational Dropout Sparsifies Deep Neural Networks (ICML 2017 · similarity 0.89)
- A Theoretically Grounded Application of Dropout in Recurrent Neural Networks (NeurIPS 2016 · similarity 0.89)
- 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.