Sally thinks Alex shouldn't have used a decision tree with 2 binary features. Instead, she thinks it would be best to use logistic regression with 16 real-valued features in addition to a bias term. Sally overherd Alex talking about this cool concept called PAC learning and she too would like to use it to analyze her method. She rst trains her logistic regression model on N examples to obtain a training error ^R . What is the the upper bound on the true error R in terms of ^R , , and N. You may use big-O notation.