Small batches are noise-limited.
Adding examples reduces gradient variance and makes each optimizer step more reliable.
Gradient noise creates a critical batch size: below it, larger batches improve update quality; above it, extra examples mostly duplicate information. Explore that bend directly.
Adding examples reduces gradient variance and makes each optimizer step more reliable.
Near the gradient noise scale, statistical efficiency begins to saturate.
More workers can shorten wall time only if communication and stragglers stay controlled.