Results for "dropout"

Randomly zeroing activations during training to reduce co-adaptation and overfitting.

Dropout is like a team sport where not every player is on the field all the time. In training a neural network, dropout randomly turns off some neurons, so the network learns to rely on different parts of itself instead of just a few. This helps prevent the network from becoming too specialized o...