Training & Learning

Underfitting

When a model is too simple or undertrained to capture the real pattern, so it performs poorly even on its training data.

Underfitting is the opposite problem to overfitting: the model has not learned enough to do the job, often because it is too small, trained too briefly, or given features that do not carry enough signal. The telltale sign is weak performance even on the data it was trained on. The fix is usually a more capable model, better inputs, or more training.