15
Information Retrieval in High Dimensional Data 1 Information Retrieval in High Dimensional Data Wintersemester 2011213 Prof. Dr. M. Kleinsteuber and Dipl. Math. M. Seibert, Geometric Optimization and Machine Learning Group, TU München

Information Retrieval in High Dimensional Data 1 Wintersemester 2011213 Prof. Dr. M. Kleinsteuber and Dipl. Math. M. Seibert, Geometric Optimization and

Embed Size (px)

Citation preview

Information Retrieval in High Dimensional Data 1

Information Retrieval in High Dimensional Data

Wintersemester 2011213

Prof. Dr. M. Kleinsteuber and Dipl. Math. M. Seibert,Geometric Optimization and Machine Learning Group,

TU München

A test: Find this person in the audience:

Information Retrieval in High Dimensional Data 2

How do we extract/store the picture‘s information?

3Information Retrieval in High Dimensional Data

4

Where would you go for a 12 months stay? Analyze the following data:

Dataset 1

Information Retrieval in High Dimensional Data

5

Where to go for a 12months stay? Analyze the following data:

Dataset 2

Information Retrieval in High Dimensional Data

6

Where to go for a 12months stay? Analyze the following data:

Dataset 3

Information Retrieval in High Dimensional Data

Dataset 1 (Porto)

Dataset 2 (Honululu)

Dataset 3 (Canberra)

How do we extract information?

Is it possible to divide simply into „good“ and „bad“ climate?

Is it possible to visualize climate-relatedness of cities?

7Information Retrieval in High Dimensional Data

More examples

8Information Retrieval in High Dimensional Data

Speech Recognition

Text Classification

Image Analysis Recognize digits/faces

Sound Separation

Data Visualization

In this course:

9Information Retrieval in High Dimensional Data

No Support Vector Machines

No Regression

No Factor Analysis

No Random Projection

No Neural Networks

No Hidden Markov Models

No Bayes Classifier

No Self Organizing Maps

.....

Reference: I. Fodor: A survey of dimension reduction techniques, Technical Report, Berkeley 2002.

Get in touch with some of the tools!

INSTEAD: Outline of the course:

1. Curse of Dimensionality

2. Statistical Decision Making

3. Principal Component Analysis

4. Linear Discriminant Analysis

5. Independent Component Analysis

6. Multidimensional Scaling

7. Isomap vs. Local Linear Embedding

8. Christmas9. Kernel PCA

10. Robust PCA

11. Sparsity and Morphological Component Analysis

Computer Vision 10

Literature:

J. Izenman. Modern Multivariate Statistical Techniques. Springer 2008.

J.A. Lee, M. Verleysen: Nonlinear Dimensionality Reduction, Springer 2007.

T. Hastie, R. Tibshirani, J. Friedman. The elements of statistical Learning, Springer 2009.

Papers (will be provided when appropriate)

Information Retrieval in High Dimensional Data 11

GOAL

GOAL

Data Analysis

Books/Papers/Internet...

mk Studis

Communicate ContentsCommunicate Contents

Give feedback/Ask questionsGive feedback/Ask questions

work indepently

work indepently

Studismk

Accept Methods Be interested Be independent Ask questions Give feedback

Choose methods Choose topics Address the questions Accept Feedback

Have fun!

Structure of Course

Information Retrieval in High Dimensional Data 14

Lecture 2 + Tutorials 2 (M. Seibert and I) (4 assignments+1 Poster Session)

LABCOURSE (Matlab Programming/Discussion and reading group/Postersession/etc.) 3

Examination: assignments required (max. 5 x 20 pts) 33%30 mins oral examination 66% (up to two persons per exam)

Questions?

Information Retrieval in High Dimensional Data 15