What is 'Batch Size'?

What is 'Batch Size'?

Approach: Define the number of training examples used in one iteration of weight updates.

The number of training examples utilized in one iteration. Smaller batch sizes offer a regularizing effect and require less memory; larger batch sizes provide more stable gradient estimates.

Follow-up: How does batch size affect training speed?

Key concepts: sgd, memory, training.