Fill This Form To Receive Instant Help
Fill This Form To Receive Instant Help
Homework answers / question archive / (d) Suppose you are running a learning experiment with a new algorithm for binary classification, predicting whether the input is positive or negative
(d) Suppose you are running a learning experiment with a new algorithm for binary classification, predicting whether the input is positive or negative. The data set you use consists of 100 positive and 100 negative examples. You compare your algorithm to a simple majority classifier. (A majority classifier always outputs the class category that forms the majority of the training set, regardless of the input.) i. You plan to use leave-one-out cross validation (see Section 18.4 of Russell and Norvig "Artificial Intelligence: a Modern Approach" ) for evaluation and you expect the ma- jority classifier to score about 50% accuracy on leave-one-out cross validation, but to your surprise, it scores zero every time. Explain why it happens. ii. Describe how the performance of the majority classifier will change as the parameter k in k-fold cross validation is reduced from 200 to 10. Do you ever expect to see an accuracy greater than 50%?
