What is Batch Size in Deep Learning and how does it affect training stability and generalization?

Batch size is the number of training samples processed before updating model weights, and it significantly impacts convergence speed, stability, and generalization.

Batch size is a core hyperparameter in deep learning that controls how many samples are used to compute gradients before updating model weights.

Core idea: Instead of updating weights after every sample, gradients are averaged over a batch.

Types:

Why batch size matters: