Techniques & Methods
Validation
Validation involves assessing a model on a held-out validation dataset during training to monitor generalization performance and tune hyperparameters. It provides an unbiased signal about whether the model is learning meaningful patterns or overfitting.
Cross-validation (k-fold) is a robust validation approach that uses multiple train-validation splits. Separate test sets are used only once after training concludes to give the final unbiased performance estimate.
Authority Links
Related Terms
Miscellaneous
Validation Data
A held-out data split used during training to tune hyperparameters and monitor generalization.
Core Concepts
Overfitting
Model learns detail and noise in training data too thoroughly, reducing generalization.
Techniques & Methods
Evaluation Metrics
Quantitative measures used to assess how well an AI model performs on a task.
Techniques & Methods
Training
Teaching a model to make accurate predictions by exposing it to large datasets.