# Sparse PCA Say you have a bunch of documents. Each document is represented as a long sparse vector .  We then have a data matrix of all document vectors:  Given this we can compute a correlation matrix (how many times does word i appear with word j in a document):       --- Date: 20240524 Links to: Tags: References: * [Megasthenis Asteris: A Framework for Sparse PCA - YouTube](https://www.youtube.com/watch?v=uCzGi6zLu6s)