10 (1 point) If a validation data is used to mitigate overfitting in decision tree construction, which one is a correct description of the approach? Question 10 options: A validation data is used to measure performance, not to mitigate overfitting Stop training once a reduction in validation accuracy is observed Continue the training until training accuracy no longer improves Continue the training as long as either training or validation accuracy improves Question 11 (1 point) In general, an overfitted model doesn't generalize well. What does it mean? Question 11 options: Its performance on unseen data is lower It hasn't learned sufficient patterns from the training data It can't be tuned or improved further It can't be applied for other domains Question 12 (1 point) Which one of the following is NOT true about tree ensemble learning? Question 12 options: It is easier to interpret tree ensembles than a single decision tree It is similar in concept to experts reaching at a decision collectively Tree ensemble almost always overfits more than a single decision tree Many trees working together is better in performance than a single decision tree Question 13 (1 point) A tree ensemble method that draws bootstrap samples (random samples with replacement) from the data and averages the results to reduce the variance is: Question 13 options: Bagging random forest Boosting Gradient descent