(a) How much information does it take to describe a two-input perceptron, sgn(w" (1,x1,x2]"))? The classical description uses a vector of three real-valued parameters, w=(wo, w1,W2)'. But the perceptron's decision boundary is a line, which can be uniquely specified with just two parameters. Chip says: I claim a perceptron can be described with less information than three real numbers. Here's how I would do it with just two real values: set si = wi/wo and s2 = w2/wo. From the description ($1,82) T, I can construct a weight vector (1,$1,82) that describes a line with the same slope and intercept as the one described by w. So it should behave exactly the same as w for all inputs." Dale replies: "I claim a perceptron requires more than just two real numbers to describe. A line is not an inequality. Consider the case where wo is negative. How will that affect the results of your transformation?" Whose claim is correct, and why? (Don't worry about the case where wo = 0.) (b) Consider a binary input variable with odd number of bits, x (-1,+1}", for some D=2N+1. We are interested in a perceptron that fires +1 whenever a majority of bits in the input variable are positive, and -1 otherwise. Give the perceptron parameters weR+ (you do not need to train it using the perceptron algorithm...)