Principal Component Analysis (PCA)

Pattern recognition algorithm has many usage in many field especially in robot. And relevantly used algorithm is Principal component analysis(PCA), and I introduce basic concept of PCA.

Basic function of PCA is that reduce dimension of big size data which has high dimension. Of course, High dimension data is represent all of data. but it is useless to use them all. Because, they have some kind of relationship, and don’t need them all. So PCA delete redundant data set, and reduce dimension of data. For example, represent point as (1,1) or (2,2) in two dimension space, need two data. But if we know the one axis (1,1), then we can make them by one point data in one dimension space.

To find axis for some data, we use singular value decomposition(SVD) or covariance matrix.

Another characterist of PCA, they use distribution characteristic of data. This makes them, enhance the classification of data, reduce noise etc. This make reduce error.

Basic tutorial of PCA and well described paper is “Smith L., A tutorial on principal component analysis.” If you interest about PCA you can find this paper.

Leave a Comment

Your email address will not be published. Required fields are marked *