Q3 In this question, you will propose versions of basic k-means discussed in the class [from PRML] to suit each of the following requirements. For each of the requirements, you have to suggest a suitable variation of basick- means version, choice of initial cluster centers, and number of centers, convergence criteria along with your comment on convergence & performance tradeoffs. a. Given the performance of students over the past 20 years in the first year courses [say'd' courses], in BITS, find a suitable cutoff to regrade the students around k groups. It is expected that the clusters should give directions on students' academic inclinations which will help the university to propose specializations. Ignore academic aspects related to evaluation from consideration. b. C. Assume that you have large data to be clustered around k clusters in a personal laptop. This has to be done in a few mins. Assume the data is in your disk and does not fit in primary memory entirely. Assume that the data arrives in streams. Consider for example the click streams that google news page receives for example / or the orders that amazon continuously receives from across the globe. It is necessary to learn the evolution of customer behavior by monitoring how the clusters evolve over