Identify instance spaces, features and labels for a new learning problem, and identify the hypothesis spaces for various learning settings.
Implement decision trees and the decision tree learning algorithm.
Illustrate online learning by example, and formally define mistake-bound learning.
Formally define Occam's razor in the context of the probably approximately correct model for learning, and apply this model for simple hypothesis classes.
Define Bayesian criteria for learning, in particular, the maximum a posterior and the maximum likelihood criteria, and apply them to the naive Bayes and logistic regression classifiers.
Define loss functions corresponding to SVM, logistic regression and a simple neural network.
Implement several supervised learning algorithms for linear classifiers in a language of their choice, including the Perceptron and SVM.