Problem 2. (60+10/60) On classifier fusion. Assume there are two classifiers (L1, L2), performing classification tasks on three objects (, 2, 3). Assume there are 30 samples in the training data for each category. Following is the confusion matrices generated from the training data LI 002 ! 20 24 003 21 O) 25 19 In the confusion matrix, each row represents the ground truth label and each column represents the label given by the classifier. (a) (10/10) Suppose for a certain testing sample, classifier 1 assigns it to , and classifier 2 assigns it to o2, what is the assigned label after fusing both classifiers using Nave Bayes rule? (Note that by dividing each matrix element by the sum of all elements in the same row, you can arrive at a probability value that indicates the likelihood. Nave Bayes assumes the classifiers are independent from each other) (b) (30/20) Derive a lookup table that includes the fused result from all possible combinations of labels from the two classifiers LI, L2 1, 1 1,2 1, 3 2, 1 2,2 2, 3 3,1 3,2 3, 3 Fused label (c) (Only 571) (+10/10) Provide an analytical study on how Nave Bayes-based fusion approaches find the fused label according to the maximum posterior probability. Problem 2. (60+10/60) On classifier fusion. Assume there are two classifiers (L1, L2), performing classification tasks on three objects (, 2, 3). Assume there are 30 samples in the training data for each category. Following is the confusion matrices generated from the training data LI 002 ! 20 24 003 21 O) 25 19 In the confusion matrix, each row represents the ground truth label and each column represents the label given by the classifier. (a) (10/10) Suppose for a certain testing sample, classifier 1 assigns it to , and classifier 2 assigns it to o2, what is the assigned label after fusing both classifiers using Nave Bayes rule? (Note that by dividing each matrix element by the sum of all elements in the same row, you can arrive at a probability value that indicates the likelihood. Nave Bayes assumes the classifiers are independent from each other) (b) (30/20) Derive a lookup table that includes the fused result from all possible combinations of labels from the two classifiers LI, L2 1, 1 1,2 1, 3 2, 1 2,2 2, 3 3,1 3,2 3, 3 Fused label (c) (Only 571) (+10/10) Provide an analytical study on how Nave Bayes-based fusion approaches find the fused label according to the maximum posterior probability