1. The following examples from a two-class classification problem are given: Class1: [2 2]T, [3 5]T; Class 2 [1 3]T, [-1 -0.5]T Starting with an augmented weight vector, [1 1 1]T, determine a solution vector for above data using the perceptron learning rule. Show first five steps of weight vector updating. 2. Suppose you are given a collection of weak learners where each learner is able to operate with 60% accuracy. You combine seven of these learners with a majority rule for final output. What will be the accuracy of the ensemble? 3. Consider the following eight records; each record is described by two quantitative attributes: A = (2, 10)t, B = (2, 5)t, C = (8, 4)t, D = (5, 8)t, E = (7, 5)t, F = (6, 4)t G = (1, 2)t, H = (4, 9)t. Let records “A”, “B”, “G”, and “H” be from class 1 and the remaining four records from class 2. Using this information, construct the Fisher’s linear discriminant function for this problem and determine the class label for the point M = (3, 3)t. 4. Consider the following six examples with three attributes: Example # Color Shape Size Class 1 Red Square Big + 2 Blue Square Big + 3 Red Round Small - 4 Green Square Small - 5 Red Round Big + 6 Green Square Big -
Determine the best attribute for root node of a decision tree classifier for above data. 5. Let , , and be four items for clustering. Consider the following three partitions: A. B. C. . Determine the partition favored by the sum-of-square-error (SSE) clustering criterion. 6. Consider the eight records of Exercise #3 without their class labels. Apply complete link clustering to this data and produce the dendrogram. This exercise must be done by hand without clustering software