PASCAL Bootcamp in Machine Learning

Lectures on Clustering

Description

These lectures give an introduction to data clustering: we discuss a few algorithms, but also look at theoretical questions related to clustering. The first two lectures are devoted to spectral clustering: graph Laplacians and their properties, spectral clustering algorithms, mathematical derivations of the algorithms, and some implementation issues. Moreover, we discuss the related modularity approach for detecting communities in networks. The third lecture is devoted to the very general question "what clustering is". We try to look at clustering from different angles, discuss different definitions of clustering, and look into theoretical foundations of clustering in general. In the last lecture we work on the question how the number of clusters should be defined. The focus is on two popular approaches: the gap statistics and the stability approach.

Slides
0:00	Clustering
0:41	Overview
1:46	Cliustering: Intuition
1:48	What is clustering, intuitively?
4:47	Example: Clustering gene expression data
5:57	Example: Social networks
6:52	Example: Image segmentation
7:16	Example: Genetic distances between mammals
8:15	Two classes of algorithms
8:17	The standard algorithm for "flat" clustering: K-means
8:30	Spectral clustering
8:46	Spectral clustering on one slide
10:50	Graph notation
14:36	Unnormalized graph Laplacian - 1
16:59	Graph notation
17:06	Unnormalized graph Laplacian - 1
18:01	Unnormalized graph Laplacian - 2
21:57	Unnormalized graph Laplacian - 3
26:02	Normalized graph Laplacians - 1
28:16	Normalized graph Laplacians - 2
28:17	Spectral clustering - main algorithms
33:32	Toy example with three clusters - 1
34:35	Toy example with three clusters - 2
35:50	Toy example with three clusters - 3
37:13	Toy example with three clusters - 4
42:47	Toy example with three clusters - 5
43:47	Toy example with three clusters - 6
43:49	Graph cut explanation of spectral clustering - 1
50:02	Graph cut explanation of spectral clustering - 2

Lecture rating

People found this lecture:
Worth seeing
because it is:
Valuable and informative
Well presented
Easily understandable
Acceptably recorded
You need to login to cast your vote.

Report a problem or upload files

If you have found a problem with this lecture or would like to send us extra material, articles, exercises, etc., please use our ticket system to describe your request and upload the data.
Enter your e-mail into the 'Cc' field, and we will keep you updated with your request's status.

Related content

• See Also
• Personal history
• More by author

Visitors who watched this lecture also watched...

[syn]  9890 views, 03:02:12 Basics of probability and statistics Mikaela Keller 27 comments	[syn]  537 views, 03:39:05 Clustering - An overview Marina Meila
[syn]  6881 views, 04:59:19 Machine Learning, Probability and Graphical Models Sam Roweis 13 comments	invited talk [syn]  1276 views, 01:17:48 Introduction to Machine Learning Isabelle Guyon
811 views, 01:08:24 PANEL: Experiences in research, teaching, and applications of ML Joaquin Quiñonero Candela, Colin de la Higuera, Isabelle Guyon, José Balcázar, Mark Girolami, Mikaela Keller, Ulrike von Luxburg	[syn]  3966 views, 03:54:31 Support Vector Machines Chih-Jen Lin 8 comments
[syn]  4163 views, 01:00:47 Gaussian Process Basics David MacKay	382 views, 02:37:54 Theory and Applications of Kernel Space Florence d'Alché

 Watch videos:   (click on thumbnail to launch)

Part 1: Spectral clustering 0:58:33 !NOW PLAYING	Part 2: Spectral clustering 0:51:05	Part 3: What is clustering, after all? 0:51:00
Part 4: Selecting the number of clusters 0:43:42

SEE ALSO:

Download slides: luxburg_clustering_lectures.pdf (3.0 MB)

Launch in a standalone WM Player

Switch to Windows Media Player

---

Streaming Video Help

Windows Media Player Firefox Plugin - Download

Reviews and comments:

1 Ryan, September 28, 2007 at 2:59 a.m.:

Please post these lectures in wmv format. Thanks

---

2 naveen, February 22, 2008 at 11:45 a.m.:

The presentation is very good, but maybe the lectures could use a bit more detailed explanations for the benefit of less gifted students.

---

Write your own review or comment:

Your name:

Comment: