Under the online learning framework, we want to learn a concept from concept class C, where C is the set of 2-dimensional thresholds (a line separator). We do not make any assumption about the location of the example points with respect to C (i.e., they can be anywhere in the 2d-plane). Specify the correct statement that best describes the mistake bound (the number of mistakes in the adversarial setting) for learning C when the input has length m, that is, we receive m examples. a. None of the other answers are correct. b. The best online learning algorithm makes 3 mistakes. C. The best online learning algorithm makes log2(m) mistakes. d. The best online learning algorithm makes 7 mistakes. e. The best online learning algorithm makes m mistakes.

Under the online learning framework, we want to learn a concept from concept class C, where C is the set of 2-dimensional thresholds (a line separator). We do not make any assumption about the location of the example points with respect to C (i.e., they can be anywhere in the 2d-plane). Specify the correct statement that best describes the mistake bound (the number of mistakes in the adversarial setting) for learning C when the input has length m, that is, we receive m examples. a. None of the other answers are correct. b. The best online learning algorithm makes 3 mistakes. C. The best online learning algorithm makes log2(m) mistakes. d. The best online learning algorithm makes 7 mistakes. e. The best online learning algorithm makes m mistakes.