Crashes when a feature variance is 0 about pca

[targos]

I'm working on the project again. I'll have a look at this.

[loretoparisi]

any news?

[targos]

What should we do when the standard deviation is 0 ? Subtract the mean without scaling? I'm going to check what R does.

[targos]

R also throws an error if the dataset cannot be scaled because of a constant column. Is there precedent in other libraries / languages? What do they do in this case?

[zyxue]

I think if there are multiple features, the one with no variance should just be ignored as it doesn't contribute anything to PCA.