Dropout
A regularization technique that randomly deactivates neurons during training to prevent overfitting.
In italiano: DropoutDuring each training iteration, dropout temporarily removes a random percentage of neurons. This forces the network to learn more robust and generalizable representations.
Examples
- Dropout with rate 0.5 in dense layers
- Reducing overfitting in deep networks
- Improving generalization