Explain 'Dropout' as an ensemble method.

Explain 'Dropout' as an ensemble method.

Approach: Describe how dropout creates a virtual ensemble of sub-networks.

Dropout can be viewed as an efficient way to train an ensemble of many different sub-networks. During testing, the 'average' of these networks is used, which improves generalization.

Follow-up: How does it compare to standard bagging?

Key concepts: sub-networks, bagging, regularization.