Back to Glossary

Batch / Epoch

Data / Training / Evaluation

Training subsets and full dataset passes.


A batch is a subset of training data processed at once, while an epoch refers to one complete pass over the dataset.

  • Purpose: Reduce memory load and stabilize learning.
  • Note: Multiple epochs are required for convergence.