ICA - Independent Component Analysis (ICA) Sergey Kirshner...

Info iconThis preview shows pages 1–13. Sign up to view the full content.

View Full Document Right Arrow Icon
Independent Component Analysis (ICA) Sergey Kirshner Department of Statistics Purdue University November 11, 2010 Guest Lecture for CS 573 (Data Mining) Acknowledgements: Barnabas Poczos (CMU) contributed many of the slides
Background image of page 1

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon
Independent Component Analysis (ICA, The Cocktail Party Problem) Estimation y = W x Sources Observation x = A s s Mixing 1 2 3 d http://www.cis.hut.fi/projects/ica/cocktail/cocktail_en.cgi
Background image of page 2
Contents ICA – Formulation – Applications – Extensions ISA Approaches – FastICA – Rank based approaches Schweizer-Wolff ICA (SWICA)
Background image of page 3

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon
Independent Component Analysis Independent signals Mixtures ICA estimation WA=
Background image of page 4
Some ICA applications Medical signal processing – fMRI, ECG, EEG Brain computer interfaces Image denoising Modeling of the hippocampus, place cells Modeling of the visual cortex Microarray data processing Decomposing the spectra of galaxies Blind deconvolution Feature extraction Face recognition Time series analysis Financial applications Clustering Classification
Background image of page 5

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon
ICA applications: removing artifacts from electroencephalography (EEG) EEG ~ Neural cocktail party Severe contamination of EEG activity by eye movements, blinks, muscle, heart, ECG artifact vessel pulse electrode noise line noise, alternating current (60 Hz) ICA can effectively detect, separate and remove activity in EEG records from a wide variety of artifactual sources. [e.g., Makeig, Jung, Bell, Ghahremanii, Sejnowski 97] Scalp weights provide evidence for the components' physiological origins.
Background image of page 6
[ Jung et al 00 ]
Background image of page 7

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon
Summed projections of selected components [ Jung et al 00 ]
Background image of page 8
[ Jung et al 00 ]
Background image of page 9

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon
ICA for Microarray data processing X T = = * S A X T 2 R M £ N M = number of experiments N = number of genes s k a k Assumption : • each experiment is a mixture of independent expression modes ( s 1 ,... s K ). • some of these modes (e.g. s k ) can be related to the difference between the classes. a k correlates with the class labels
Background image of page 10
ICA for Microarray data processing (Schachtner at al, ICA07) Brest Cancer Data set 9 th column of A: Class 1, weak metastasis Class 2, strong metastasis M=14 Experiments N=22283 genes 2 classes |Corr( a 9 , d )|=0.89, where d is the vector of class labels
Background image of page 11

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon
ICA for Microarray data processing (Schachtner at al, ICA07) Leukemia Data set ALL-B AML M=38 Experiments N=5000 genes 3 classes: ALL-B, ALL-T, AML ALL-T ALL-B AML ALL-T
Background image of page 12
Image of page 13
This is the end of the preview. Sign up to access the rest of the document.

This note was uploaded on 03/13/2012 for the course CS 573 taught by Professor Staff during the Fall '08 term at Purdue University-West Lafayette.

Page1 / 58

ICA - Independent Component Analysis (ICA) Sergey Kirshner...

This preview shows document pages 1 - 13. Sign up to view the full document.

View Full Document Right Arrow Icon
Ask a homework question - tutors are online