Mini-Batch
A mini-batch is a small subset of training examples used to estimate the gradient in one iteration of SGD-style optimization.
Batching improves hardware efficiency and gradient stability versus pure single-example SGD.
A mini-batch is a small subset of training examples used to estimate the gradient in one iteration of SGD-style optimization.
Batching improves hardware efficiency and gradient stability versus pure single-example SGD.