Description
Introduction to Machine Learning – KNN, K-means, HAG version 2
1 Introduction
In this assignment you have three short tasks. They are on k-nearest neighbor (KNN) algorithm, k-means clustering, and hierarchical agglomerative clustering (HAC). All tasks are independent from each other and have different data for you to work on. All related files can be downloaded from http://user.ceng. metu.edu.tr/~artun/ceng499/ceng499_hw2.zip. Using the lecture materials while doing the homework is strongly recommended. First part is covered in kNN-intro_to_ML.pdf, the second part is covered in k-means.pdf, and the third part is covered in hac.pdf [1]. At each part of this homework, you are going to fill the empty functions given to you. To understand how we expect them to work, you can read their descriptions or inspect the mini testers given with them. After filling those functions, you are expected to use them to complete the tasks. You can write your additional code inside mini testers or in a different file.
2 Part 1: K-Nearest Neighbor
2.1 Dataset The dataset is created by modifying and introducing some noise to the Iris Dataset [2], so you need to use the version we provided. It resides in the folder hw2_data/knn. The dataset is labeled (since the task is supervised learning) and have separate train and test sets. It is in the form of saved numpy array and you can load it by: train_data = np . load( ’hw2_data/knn/train_data . npy ’ ) train_labels = np . load( ’hw2_data/knn/ train_labels . npy ’ ) test_data = np . load( ’hw2_data/knn/test_data . npy ’ ) test_labels = np . load( ’hw2_data/knn/ test_labels . npy ’ )
2.2 Task
You are expected to fill the empty functions and use k-fold cross-validation on training data to find a suitable k for the KNN algorithm and test your KNN algorithm using that k on test dataset and report your accuracy. You should do:
1
