Instructions
• This assignment should be completed individually.
• Do not look at solutions to this assignment or related ones on the Internet.
• The files related to the assignment are present in lab4-rollno.zip folder. Extract it and upload it on moodle in the same .zip format after completion and after replacing the string “rollno” with your actual roll number. For example, if you roll number is 00405036, then single zip folder that you will upload will be named “lab4-00405036.zip”. Also collate all the CS337 based theory solutions into ONE pdf file named answers.pdf. Include answers.pdf inside the zip folder mentioned above and submit the zip folder.
• Answers to all subjective questions need to be placed in single pdf answers.pdf including all plots and figures and uploaded.
• Only add/modify code between TODO and END TODO unless specified otherwise. You must not import any additional libraries.
1 Clustering
1.1 CS 335 KMeans Implementation
1
Figure 1: Concentric Circles Dataset
2 Kernel design and Kernelized clustering
2.1 CS 337: Proving kernel validity
Prove that the function Kσ : Rn ×Rn → R defined as is a valid Kernel.
2.2 CS 337: Simple Kernel Design
Figure 1 shows a zero-centered dataset where blue points have label 0 and red points have label 1.
Each blue point is at a distance ) from the origin and each red point is at a distance ) from the origin.
You will implement KMeans using the kernel function you identify here in your programming assignment by completing the functions fit, predict of Kmeans Kernel class. First you have to complete make zero centered helper function which makes any concentric circles dataset zero centered.
2
