Underfitting occurs when a statistical model or machine learning algorithm cannot adequately capture the underlying structure of the data. It typically happens when the model is too simple (low complexity) to learn the patterns, resulting in high bias and poor performance on both training and test data.
A core concept in statistical modeling and the Bias-Variance Tradeoff.
A common pitfall in early-stage modeling, usually addressed by increasing model size, adding features, or training longer.