Random projection

Intro

More computationally efficient than PCA

Random projection chooses a random line. It wotks for large dims.

Implementation

Independent Component Analysis

Assumes features are mixtures of independent sources.

In this image, you go to a concert with three sounds intermingled. ICA will allow you to fetch original noise/

Algorithm

##Applications ICA is used in EEG and FMRI data

Ref

"Independent component analysis: algorithms and applications"(http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.322.679&rep=rep1&type=pdf)