AI Termcirca 2014· Added May 30, 2026
Dropout (in Machine Learning)
Dropout is a regularization technique used to prevent overfitting in neural networks by randomly setting portions of neurons to zero during training.
Dropout is a widely used technique in deep learning that addresses overfitting by preventing complex co-adaptations on training data. During each iteration of training, dropout randomly sets a fraction of the neurons' outputs to zero. This mimics ensemble methods as different subnetworks are trained each time, making the final model more robust and reducing the risk of dependency on specific neurons or interactions between them.
Examples
- Using dropout layers in convolutional neural networks (CNNs).
- Implementing 50% dropout rate during training phases.
Common misconceptions
- Dropout improves performance by itself; it mainly helps avoid overfitting not directly boosting accuracy.
- It's only useful for deep networks; while prevalent there, it applies broadly across models with large parameter spaces.
Related terms
Want more like this?
Open the full library
Fresh AI mastery content every 2 hours.