A guide to sparse PCA
PCA is a popular tool for exploratory data analysis and dimension reduction, especially in the high-dimensional setting. To improve interpretability, several PCA methods generating sparse solutions have been proposed. Solving the sparse SCA problem is intractable, given its combinatorial nature. In this seminar, different sparse PCA methods are analysed, focusing on the optimisation criteria used to achieve sparseness. Practical issues are discussed, such as the misconception that equivalent PCA formulations remain equivalent under sparsity. Finally, an extension to the problem of sparse Simultaneous Component Analysis is presented.
Zoom link: https://eur-nl.zoom.us/j/95156986071?pwd=NmhWa2pYbWRoL3F1SWtxcElGZUhOQT09
Meeting ID: 951 5698 6071