Github user mengxr commented on the pull request: https://github.com/apache/spark/pull/964#issuecomment-45990973 @vrilleup This implementation looks good to me and thanks for the experiments! Besides the inline comments, we should think when to switch from ARPACK to dense SVD. ARPACK needs O(k) iterations and O(k n) storage on master, while SVD needs 1 iteration and O(n^2) storage on master and each worker. For small `n` like 100 or medium `n` and `k` close to `n`, dense SVD may have advantage. Could you test some small `n`s and see which is faster?
--- If your project is set up for it, you can reply to this email and have your reply appear on GitHub as well. If your project does not have this feature enabled and wishes so, or if the feature is enabled but not working, please contact infrastructure at infrastruct...@apache.org or file a JIRA ticket with INFRA. ---