// Note that equation (2) says 1 for the minimum error, but the code says 2.0.
// And it says L2 norm, but really, the code does absolute value even for
// L2 error norm (it uses a square root formula for L1 norm).
// Note that equation (2) says 1 for the minimum error, but the code says 2.0.
// And it says L2 norm, but really, the code does absolute value even for
// L2 error norm (it uses a square root formula for L1 norm).