Footnote

This statement is a bit sloppy. If the number n of examples is small, the covariance for independent features will not turn out to be exactly zero. In fact, it might not even be very small. However, for a large number of examples, we can be confident that the magnitude of the covariance will be small compared to its maximum possible value, which is the product of the standard deviation of Feature i and the standard deviation of Feature j.

Back