I working on a gaussian process and I'm running into a problem with computing the log likelihood when the matrix gets large(300+) and the parameter guesses suck(but are valid). I'm doing this in C++ ...