Mini Batch - AshokBhat/ml GitHub Wiki

Description

• Instead of an entire batch, a small randomly selected subset is used in each training iteration
• The batch size is usually between 10 and 1,000.

Pros and cons

• Pros - More efficient
• Cons - Staggerred path to minima

See also

• [Batch]] ](/AshokBhat/ml/wiki/[[Batch-size) | Mini-Batch
• [Iteration]] ](/AshokBhat/ml/wiki/[[Forward-pass) | Backpropagation