Algorithm for High Dimensional Principal Component Analysis

Alexander Ilin and Tapani Raiko

This page provides an algorithm that is efficient when the dimensionality of the data d is high compared to the number of principal components c needed.

Here is a short description of the algorithm (PDF) and the source code (Matlab function).

This algorithm was presented in [1], please give a citation if you find this useful. The paper and provided Matlab package also includes extensions such as variational Bayesian treatment of missing values.

References

[1] A. Ilin and T. Raiko. Practical approaches to principal component analysis in the presence of missing values. Journal of Machine Learning Research, volume 11, pages 1957-2000, July 2010.