📝 Paper: Polis: Scaling Deliberation by Mapping High Dimensional Opinion Spaces (2021)
The particular algorithm used to perform PCA is the power iteration method (Roweis, 1997). This method allows us to use the previous principal component eigenvectors as the starting point for further iterations of the method when there are new votes to process, allowing the method to converge very quickly, and with reduced computational load.
📝 Paper: EM Algorithms for PCA and SPCA (1997)
Candidate Implementations