Definition
A specific technique for preventing a neural network from overfitting by randomly dropping out a fraction of neurons during each training step.
In short: A specific technique for preventing a neural network from overfitting by randomly dropping out a fraction of neurons during each training step.