This repository contains code for my Masters Thesis "Principal component analysis as optimization on Riemanian manifolds".
To quote my summary of the thesis itself: "This master thesis presents a modified Manifold Sparse PCA algorithm. The derived algorithm uses smooth manifold optimization to find sparse representations of the empirical covariance matrix. The motivation behind this algorithm was a lack of interpretability of the simple PCA. In the presented approach, the sparse representation of the initial matrix allows for an interpretation of the principal components. Numerical simulations presented in this thesis are in line with theoretical expectations. At the end of the thesis, further research perspectives are presented, i.e. we propose some approaches to improve the algorithm itself, as well as its implementation."
Thesis itself was graded 5.0 by all the reviewers, and can be found as a separate file in this repository (although it is in Polish).