Learning Curves

A learning curve shows how model performance changes as you add more training data. The gap between training and validation accuracy reveals whether your model is underfitting or overfitting.

A straight line can't capture this curved pattern — it misses the shape entirely. Both learning curves plateau low because the model is fundamentally too simple.

The model memorizes training data (low train error) but the wiggly curve misses validation points. More training data would help tame the overfitting.

With a simple model, the gap is flat — both curves plateau low. More data won't help. You need a more complex model to improve.

Good fit! The model balances complexity and generalization — train and validation accuracies are both high with a small gap.

Your model is overfitting — there's a large gap between train and validation. Simplify the model (regularization, fewer parameters) or collect more training data.