AI Engineering Degree Practice Exam 2025 - Free AI Engineering Practice Questions and Study Guide

The statement regarding performing training and testing on the same dataset leading to very high out-of-sample accuracy is not true. When a model is trained and tested on the same dataset, it often results in an overfitted model, which means that the model has learned the specifics and noise of the training data too well. This leads to inflated accuracy metrics, as the model appears to perform exceptionally when evaluated on the same data it learned from.

In reality, out-of-sample accuracy is a measure of how well the model generalizes to unseen data. If the same dataset is used for both training and testing, the model's reported accuracy does not provide a valid indication of its predictive performance on new, unseen data. A true assessment of the model's performance requires rigorous testing on independent datasets to ensure that the accuracy reflects the model's ability to generalize rather than its capacity to memorize the training examples.