Although bias and variance have theoretical formulas, it is difficult to calculate their actual values. A simple way to estimate them empirically is with learning and validation curves.