Definition
A regularization technique where randomly selected neurons are temporarily ignored (set to zero) during training to prevent over-reliance on specific features and reduce overfitting.
In short: A regularization technique where randomly selected neurons are temporarily ignored (set to zero) during training to prevent over-reliance on specific features and reduce overfitting.