Training data

Training data is the labeled dataset used to teach a machine learning model patterns that generalize beyond the examples it has seen.

In supervised learning, it typically consists of paired inputs and corresponding labels that serve as ground truth.

Data is usually split into training, validation, and test sets to evaluate model performance and monitor overfitting.

The representativeness, label accuracy, scale, and diversity of training data are major drivers of generalization to new data. Any bias or error present in the data will propagate directly to the model.