Question 7

• Given two learning algorithms and and a large data set , which algorithm will produce more accurate classifiers when trained on datasets of a specified size ?
• Also not much attention.
• DELVE team has studied it for regression.
• They divide into several disjoint training sets and a single test set.
• Each algorithm is trained on each training set, and all resulting classifiers tested on the test set.
• An analysis of variance is performed that includes terms for choice of the learning algorithm, choice of the training set, and each individual test example.
• Quasi-F test is applied to determine whether the effect due to the choice of learning algorithms is significantly non-zero.