add a sparse PCA step

[topepo]

Via the VBsparsePCA package.

[juliasilge]

I'm not sure if you have specific reasons for wanting to use VBsparsePCA (Bayesian vs. not?), but there's a different implementation of PCA for sparse data in irlba that I have used a fair amount with great results. It is what is used in widyr as well.

[topepo]

I'll check it out. I saw VBsparsePCA when it was recently released and liked that it was fast and consistent with prcomp() results.

[topepo]

One difference is that VBsparsePCA applies the sparsity for the all the loadings for a predictor (so it has no affect on all components). irlba doesn't use that constraint. Since I would only need to keep the rotation matrix, it might be easy to do both (step_pca_sparse() and step_pca_sparse_bayes()?).

[EmilHvitfeldt]

There is also a sparse pca step in the MachineShop package using the {elasticnet} package

[github-actions]

This issue has been automatically locked. If you believe you have found a related problem, please file a new issue (with a reprex: https://reprex.tidyverse.org) and link to this issue.