Sometimes I just disappear.
Dropout randomly turns off some neurons during training to prevent overfitting and make neural networks more robust.
Imagine a team of students studying for a test. If they always rely on the same star student, they'll fail if that student is absent. Dropout randomly 'sends some students home' during practice so everyone learns to solve problems on their own, making the whole team stronger.
Dropout is crucial for training large neural networks without them memorizing the training data instead of learning general patterns. You encounter it in almost every modern deep learning framework, from image recognition to language models.
Many think dropout is used during both training and testing. In reality, dropout is only applied during training; during testing, all neurons are active but scaled down to compensate.
Dropout is a regularization technique where randomly selected neurons (along with their connections) are temporarily removed from the neural network during training. This prevents co-adaptation of neurons, forcing the network to learn more robust features that are useful in isolation. At test time, all neurons are used but their outputs are multiplied by the dropout probability to maintain consistency.