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 | Batch size | Mini-Batch
Iteration | Forward pass | Backpropagation

⚠️ GitHub.com Fallback ⚠️