4. The classic PAC learning model of Valiant'84 made two fundamental assumptions: 1) the distributions of the training data and testing data are the same; and 2) all instances are labeled correctly. The goal of PAC learning is to nd a hypothesis whose error rate, i.e. the probability that it misclassies a new sample, is upper bounded by e E (0, 1). Give two respective examples to illustrate that if either assumption is violated, PAC learning becomes impossible