Research map / Architectures and efficiency / RNNs and normalization
Shakeout: A New Regularized Deep Neural Network Training Scheme
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 |
|---|---|---|---|
| replace Dropout with Shakeout on hidden units | fully connected network · MNIST, small training sets | test error (%): 11.76 -> 10.81 (500 examples) | Helped |
| replace Dropout with Shakeout on hidden units | fully connected network · MNIST, large training sets | test error (%): 1.36 -> 1.35 (50000 examples) | No clear effect |
| replace Dropout with Shakeout on hidden units | CNN · MNIST, small training sets | test error (%): 6.16 -> 4.83 (500 examples) | Helped |
| replace Dropout with Shakeout on hidden units | CNN · MNIST, 8000 training examples | test error (%): 1.23 -> 1.31 | No clear effect |
| replace Dropout with Shakeout on FC layer | cuda-convnet · CIFAR-10, small training sets | test error (%): 65.34 -> 63.71 (300 examples) | Helped |
| replace Dropout with Shakeout on FC layer | cuda-convnet · CIFAR-10, 40000 training examples | test error (%): 20.50 -> 20.56 | No clear effect |
| add Dropout or Shakeout regularization versus standard BP | fully connected network · MNIST | test error (%): 13.66 -> 10.81 (500 examples) | Helped |
| increase Shakeout constant c with tau fixed (ablation) | Shakeout autoencoder · MNIST autoencoder weight sparsity | weight sparsity | Helped |
Closest papers by meaning
- Dropout with Expectation-linear Regularization (ICLR 2017 · similarity 0.90)
- Dropout Reduces Underfitting (ICML 2023 · similarity 0.89)
- Regularizing Deep Neural Networks by Noise: Its Interpretation and Optimization (NeurIPS 2017 · similarity 0.88)
- Adaptive Dropout with Rademacher Complexity Regularization (ICLR 2018 · similarity 0.88)
- Guided Dropout (AAAI 2019 · similarity 0.87)
- Jumpout : Improved Dropout for Deep Neural Networks with ReLUs (ICML 2019 · 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.