What Is a Learning Curve in Machine Learning?

Machine-Learning Learning Curve Explainer

A learning curve plots a model’s performance against a growing amount of training data or, in some settings, against training iterations. It helps reveal whether additional experience is improving both fit and generalization.

Curves for the training and validation sets are especially useful together. A persistently large gap can suggest high variance, while two weak curves that meet at a poor score can suggest high bias or inadequate features.

The shape should be interpreted with the metric, split strategy, and uncertainty across repeated samples. A jagged curve can reflect small datasets or unstable estimates rather than a meaningful change in the learning process.

Learning curves do not prescribe one automatic fix. They help decide whether to collect more data, change model capacity, improve representation, adjust regularization, or investigate labeling and distribution problems before spending more resources.

A learning curve shows how model performance changes as training experience or the amount of available data increases.

Key terms include learning curve, training score, validation score, sample size, training iteration, bias, variance and generalization gap.

Training and validation curves show whether performance converges as the model receives more experience.

A wide gap and two weak converged curves point to different kinds of investigation.

The evidence may support collecting data, changing capacity, improving features, or revisiting labels.

Uncertainty bands and repeated splits keep small fluctuations from being mistaken for durable improvement.

Explore more "Explainers"

Discover additional explainers across politics, science, business, technology, and other fields. Each explainer breaks down a complex idea into clear, everyday language—helping you better understand how major concepts, systems, and debates shape the world around us.