👉 In machine learning, a "dropout" is an operation that randomly sets some units of a model to zero. This operation helps avoid overfitting by allowing the model to learn from data points that are not representative of the target class. The term "dropout" was first introduced in the context of artificial neural networks and has since been widely used in a variety of applications, including image classification and natural language processing. In simple terms, dropout allows the network to "drop out" or ignore certain units