Well, all functions that use finite precision arithmetic deal with approximations. Errors of 1/2 ulp like trig functions are rather the exception. And machine l

Well, all functions that use finite precision arithmetic deal with approximations. Errors of 1/2 ulp like trig functions are rather the exception. And machine learning is keen on using low precision floating point type, so errors are relatively large.