I don't see the problem with PCA if it is used sensitively to identify the approximate dimensionality of "pancakes" embedded in a larger space.
I don't see the problem with PCA if it is used sensitively to identify the approximate dimensionality of "pancakes" embedded in a larger space.