In this problem, we will look at classifying protein sequences according to whether they belong to a particular protein family or not. For this task, we will use the string kernel that we discussed in class. Download the ?les hw5train.txt and hw5test.txt from the class website. These ?les contain your training and test data sets respectively. The data ?les are in ASCII text format, and each line of the ?le contains a string, which represents a protein sequence, followed by a label, which is 1 or 1, to indicate whether the protein sequence belongs to a protein family or not. Each letter in the protein sequence represents an amino acid, and thus the alphabet size is 20. Di?erent protein sequences in the ?le have di?erent length; this is not surprising because even the same protein will have di?erent lengths in di?erent species, for example, in mouse and human. Assume that the data is linearly separable by a hyperplane through the origin. Run a single pass of kernel perceptron algorithm on the training dataset to ?nd a classi?er that separates the two classes. For your kernel, use the string kernel function. Recall from class that given two strings s and t, the string kernel Kp(s; t) is the number of substrings of length p that are common to both s and t. For this problem, use p = 3 and p = 4. Write down the training and test errors of kernel perceptron for p = 3 and p = 4 on this dataset.