A decision tree is trained on a dataset with a continuous feature. How does the tree determine the best split threshold for that feature?