Top Banner
Clustering and the K-means algorithm Yihui Saw 18.304 Seminar Talk I March 6, 2013 Saturday, March 16, 13
24

Clustering and the K-means algorithm - MIT Mathematicsmath.mit.edu/~rothvoss/18.304.1PM/Presentations/1-Yihui-Talk1.pdf · Clustering and the K-means algorithm Yihui Saw 18.304 Seminar

Mar 11, 2019

Download

Documents

phamkhanh
Welcome message from author
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Page 1: Clustering and the K-means algorithm - MIT Mathematicsmath.mit.edu/~rothvoss/18.304.1PM/Presentations/1-Yihui-Talk1.pdf · Clustering and the K-means algorithm Yihui Saw 18.304 Seminar

Clustering and the K-means algorithm

Yihui Saw18.304 Seminar Talk I

March 6, 2013

Saturday, March 16, 13

Page 2: Clustering and the K-means algorithm - MIT Mathematicsmath.mit.edu/~rothvoss/18.304.1PM/Presentations/1-Yihui-Talk1.pdf · Clustering and the K-means algorithm Yihui Saw 18.304 Seminar

Clustering examples

• Customer purchase patterns

• Language family models

• Data compression

Saturday, March 16, 13

Page 3: Clustering and the K-means algorithm - MIT Mathematicsmath.mit.edu/~rothvoss/18.304.1PM/Presentations/1-Yihui-Talk1.pdf · Clustering and the K-means algorithm Yihui Saw 18.304 Seminar

Original Image

Saturday, March 16, 13

Page 4: Clustering and the K-means algorithm - MIT Mathematicsmath.mit.edu/~rothvoss/18.304.1PM/Presentations/1-Yihui-Talk1.pdf · Clustering and the K-means algorithm Yihui Saw 18.304 Seminar

2 colors

Saturday, March 16, 13

Page 5: Clustering and the K-means algorithm - MIT Mathematicsmath.mit.edu/~rothvoss/18.304.1PM/Presentations/1-Yihui-Talk1.pdf · Clustering and the K-means algorithm Yihui Saw 18.304 Seminar

4 colors

Saturday, March 16, 13

Page 6: Clustering and the K-means algorithm - MIT Mathematicsmath.mit.edu/~rothvoss/18.304.1PM/Presentations/1-Yihui-Talk1.pdf · Clustering and the K-means algorithm Yihui Saw 18.304 Seminar

8 colors

Saturday, March 16, 13

Page 7: Clustering and the K-means algorithm - MIT Mathematicsmath.mit.edu/~rothvoss/18.304.1PM/Presentations/1-Yihui-Talk1.pdf · Clustering and the K-means algorithm Yihui Saw 18.304 Seminar

The clustering problem

Saturday, March 16, 13

Page 8: Clustering and the K-means algorithm - MIT Mathematicsmath.mit.edu/~rothvoss/18.304.1PM/Presentations/1-Yihui-Talk1.pdf · Clustering and the K-means algorithm Yihui Saw 18.304 Seminar

Distance metric

Saturday, March 16, 13

Page 9: Clustering and the K-means algorithm - MIT Mathematicsmath.mit.edu/~rothvoss/18.304.1PM/Presentations/1-Yihui-Talk1.pdf · Clustering and the K-means algorithm Yihui Saw 18.304 Seminar

The cost of clustering

Saturday, March 16, 13

Page 10: Clustering and the K-means algorithm - MIT Mathematicsmath.mit.edu/~rothvoss/18.304.1PM/Presentations/1-Yihui-Talk1.pdf · Clustering and the K-means algorithm Yihui Saw 18.304 Seminar

K-means algorithm

Saturday, March 16, 13

Page 11: Clustering and the K-means algorithm - MIT Mathematicsmath.mit.edu/~rothvoss/18.304.1PM/Presentations/1-Yihui-Talk1.pdf · Clustering and the K-means algorithm Yihui Saw 18.304 Seminar

K-means algorithm

Saturday, March 16, 13

Page 12: Clustering and the K-means algorithm - MIT Mathematicsmath.mit.edu/~rothvoss/18.304.1PM/Presentations/1-Yihui-Talk1.pdf · Clustering and the K-means algorithm Yihui Saw 18.304 Seminar

K-means algorithm

Saturday, March 16, 13

Page 13: Clustering and the K-means algorithm - MIT Mathematicsmath.mit.edu/~rothvoss/18.304.1PM/Presentations/1-Yihui-Talk1.pdf · Clustering and the K-means algorithm Yihui Saw 18.304 Seminar

K-means algorithm

Saturday, March 16, 13

Page 14: Clustering and the K-means algorithm - MIT Mathematicsmath.mit.edu/~rothvoss/18.304.1PM/Presentations/1-Yihui-Talk1.pdf · Clustering and the K-means algorithm Yihui Saw 18.304 Seminar

K-means algorithm

Saturday, March 16, 13

Page 15: Clustering and the K-means algorithm - MIT Mathematicsmath.mit.edu/~rothvoss/18.304.1PM/Presentations/1-Yihui-Talk1.pdf · Clustering and the K-means algorithm Yihui Saw 18.304 Seminar

K-means algorithm

Saturday, March 16, 13

Page 16: Clustering and the K-means algorithm - MIT Mathematicsmath.mit.edu/~rothvoss/18.304.1PM/Presentations/1-Yihui-Talk1.pdf · Clustering and the K-means algorithm Yihui Saw 18.304 Seminar

Proof of convergence• Each iterative step necessarily lowers the

cost - the cost monotonically decrease

Saturday, March 16, 13

Page 17: Clustering and the K-means algorithm - MIT Mathematicsmath.mit.edu/~rothvoss/18.304.1PM/Presentations/1-Yihui-Talk1.pdf · Clustering and the K-means algorithm Yihui Saw 18.304 Seminar

Proof of convergence• Each iterative step necessarily lowers the

cost - the cost monotonically decrease

]

Saturday, March 16, 13

Page 18: Clustering and the K-means algorithm - MIT Mathematicsmath.mit.edu/~rothvoss/18.304.1PM/Presentations/1-Yihui-Talk1.pdf · Clustering and the K-means algorithm Yihui Saw 18.304 Seminar

Convergence to local minimum

Saturday, March 16, 13

Page 19: Clustering and the K-means algorithm - MIT Mathematicsmath.mit.edu/~rothvoss/18.304.1PM/Presentations/1-Yihui-Talk1.pdf · Clustering and the K-means algorithm Yihui Saw 18.304 Seminar

Convergence to local minimum

Saturday, March 16, 13

Page 20: Clustering and the K-means algorithm - MIT Mathematicsmath.mit.edu/~rothvoss/18.304.1PM/Presentations/1-Yihui-Talk1.pdf · Clustering and the K-means algorithm Yihui Saw 18.304 Seminar

Convergence to local minimum

Saturday, March 16, 13

Page 21: Clustering and the K-means algorithm - MIT Mathematicsmath.mit.edu/~rothvoss/18.304.1PM/Presentations/1-Yihui-Talk1.pdf · Clustering and the K-means algorithm Yihui Saw 18.304 Seminar

Convergence to local minimum

Saturday, March 16, 13

Page 22: Clustering and the K-means algorithm - MIT Mathematicsmath.mit.edu/~rothvoss/18.304.1PM/Presentations/1-Yihui-Talk1.pdf · Clustering and the K-means algorithm Yihui Saw 18.304 Seminar

Convergence to local minimum

Saturday, March 16, 13

Page 23: Clustering and the K-means algorithm - MIT Mathematicsmath.mit.edu/~rothvoss/18.304.1PM/Presentations/1-Yihui-Talk1.pdf · Clustering and the K-means algorithm Yihui Saw 18.304 Seminar

Convergence to local minimum

Saturday, March 16, 13

Page 24: Clustering and the K-means algorithm - MIT Mathematicsmath.mit.edu/~rothvoss/18.304.1PM/Presentations/1-Yihui-Talk1.pdf · Clustering and the K-means algorithm Yihui Saw 18.304 Seminar

Convergence to local minimum

Saturday, March 16, 13