Which of these is NOT true about PCA?

a$.$

The original data can usually be reconstructed exactly as before from the reduced dimensions

b$.$

It decomposes the data along orthogonal $($perpendicular$)$ axes and keeps the ones with the largest "spread"

c$.$

For N$-$dimensional data with more than N rows, it can return anything between $1$ to $($N$-1)$ dimensional output

d$.$

It is a technique for dimensionality reduction