Will this randomized PCA operation be completed?

Question

I am trying to perform whitening on a matrix with dimensions (100000, 1024). I need to determine if the PCA or RandomizedPCA from sklearn.decomposition can cope with a problem of this size, and if not, what is the largest number of possibilities I can use and will end in about 24 hours. Implementation is not parallelized.

According to the documentation for sklearn.decomposition.PCA , the temporary complexity of transforming a PCA is' O (n ** 3), assuming n ~ n_samples ~ n_features. "But I'm not sure what this designation means.

RandomizedPCA is supposedly more efficient because it's an approximation, but I don't know how much. The documentation is nothing complicated in time.

For reference, I converted to (100000, 2) and it takes about 2 minutes. I started converting (PCA and RandomizedPCA) to (100000, 1024), which work for about an hour. I would like to know if they will ever be finished, but if I cannot, I will just finish them in 24 hours if they still work.

Answer 1

Through experimentation, a randomized PCA appears that grows linearly and ends in just over an hour for 1024 functions.