Which ONE of the following combinations of Training, Validation, Testing data is used during the process of learning/creating the model? SELECT ONE OPTION

The process of creating and learning a machine learning model involves two key iterative steps that use specific data subsets. The training data is used to teach the model the underlying patterns by adjusting its internal parameters (e.g., weights). The validation data is used concurrently to tune the model's hyperparameters (e.g., learning rate, architecture) and prevent overfitting. The test data is strictly reserved for a final, unbiased evaluation after the model has been fully trained and tuned. Therefore, only the training and validation datasets are used during the iterative learning and creation process.

A. This is incorrect because the test data is not used during the learning/creation process; its use at this stage would lead to data leakage and an over-optimistic performance evaluation.
C. This is incorrect as it omits the validation data, which is crucial for hyperparameter tuning and preventing overfitting during model creation.
D. This is incorrect as it omits the training data, which is the fundamental dataset the model learns from. Without it, no learning can occur.

1. ISTQB® Certified Tester AI Testing (CT-AI) Syllabus v1.0 (2021)

Section 2.1.1

p. 14. "The training set is used to train the model... The validation set is used to tune the model’s hyperparameters... The test set is used to provide an unbiased evaluation of the final model."

2. Goodfellow

I.

Bengio

Y.

& Courville

A. (2016). Deep Learning. MIT Press. Chapter 5.1.3

p. 110. This text clearly separates the roles of training

validation

and test sets

specifying that the test set is not used for training or tuning.