PCA scikit-learn - ValueError: array must not contain infs or NaNs

Question

I use PCA from from sklearn.decomposition to reduce data dimension.

    pca = PCA(n_components=pca_components, whiten=True, svd_solver='full')
    y = pca.fit_transform(x)

The numpy array shape is (512, 48), dtype is float64. The minimum value in the array is 0.0, the maximum value is 1.7976931348623157e+308. The array does not contain infs or NaNs but I get an error

ValueError: array must not contain infs or NaNs

I have looke for reasons of this error and found nothing useful for me. I suppose that maybe the maxim value in the array is too large for PCA. But I still do not know how deal with it. Could you advice me something. Thanks alot in advance.

Answer 1

My guess is the issue is a combination of whiten=True and the largest value equal to the largest possible value of a float64.

From scikit-learn's PCA docs:

When True (False by default) the components_ vectors are multiplied by the square root of n_samples…

That creates an overflow issue which results in the ValueError exception.