Dropout
In plain terms
Randomly disabling neurons during training to reduce overfitting.
Going deeper
Dropout randomly 'turns off' a fraction of neurons each training step, forcing the network to learn robust, redundant features. It prevents over-reliance on specific units.
Under the hood
Acts like training an ensemble of sub-networks; disabled only during training, not inference.
Examples
Neural network regularization, deep learning, preventing co-adaptation.
Think of it like...
Like practicing a team play with random players sitting out, so no one is indispensable.