Principal Component Analysis.
0/8
completed
PCA is applied to a dataset before training a classifier. The first two PCs are kept. A colleague argues this always improves classification performance. Why might this be wrong?