University:
Whitman College
Course:
CS-350 | Foundations of Machine Learning
Academic year:

2024
Views:

336

Pages:

2
Author:

ghisa3qt

Computer Lab: K-means Clustering on Digits The data we’re looking at represents an extraction of handwritten digits, where the data was preprocessed to an 8 × 8 grid of pixels. Here’s a sample of 10 random points in the set, each visualized in the 8 × 8 grid: When we load the digits (using either handdigits.mat file for Matlab/Octave or the built-in digits file in Python), we’ll have a matrix that is 1797 × 64, and this represents the 1797 data points in R64 . The “targets” vector is something we’ll use later, but it tells us which digit the image is supposed to be. Your goal is to cluster the points into 10 clusters using k-means and show the results. Be sure to run k-means several times to get the best clustering. Here’s a partial Matlab file to get you started: load handdigits % % % % Fill in the missing command, but be sure your output for the cluster centers is the 10 x 64 matrix C for the image commands below to work. 1 % Show the 10 clusters as 8 x 8 images (You don’t need to edit this) for j=1:10 subplot(2,5,j) temp=reshape(-C(j,:),8,8); imagesc(temp’); colormap(gray); axis equal; axis off end In Python, you might use: import matplotlib.pyplot as plt import seaborn as sns; sns.set() # for plot styling import numpy as np from sklearn.cluster import KMeans from sklearn.datasets import load_digits digits = load_digits() digits.data.shape kmeans = clusters = kmeans.cluster_centers_.shape # Instantiation of KMeans # Run k-means on digits.data # This part you shouldn’t need to change- This shows the 10 cluster centers # as 8x8 blocks fig, ax = plt.subplots(2, 5, figsize=(8, 3)) centers = kmeans.cluster_centers_.reshape(10, 8, 8) for axi, center in zip(ax.flat, centers): axi.set(xticks=[], yticks=[]) axi.imshow(center, interpolation=’nearest’, cmap=plt.cm.binary) 2

Computer Lab K-means Clustering on Digits

Related Documents

Description

Related Documents

Free up your schedule!

Take 5 seconds to unlock