#out-of-core

Question 1: Single choice

You want to apply PCA to a dataset that is far too large to fit in memory. Which option is designed for this?

IncrementalPCA, fed with one mini-batch at a time via partial_fit1
PCA(svd_solver="randomized"), which approximates the first components quickly2
KernelPCA with an RBF kernel, which works with a kernel matrix instead of the data3
PCA(n_components=0.95), which keeps fewer components and so less of the data4