NIPALS returns nans

[sandcobainer]

From the documentation, I understand specifying the nComponents only works for the options.method = 'NIPALS'.

However

options.method='NIPALS' 
options.nCompNIPALS = 2

pca = new PCA(dataset, options)
embedding = pca.predict(dataset, options)

always returns arrays of NaNs. [NaN,NaN]. I've tried it with 'SVD' and there are no NaNs, however the other methods don't support nComponents. So how do you use this for dimensionality reduction?

[targos]

Hello, do you have an example dataset that shows the issue?