After moving from theoretical high energy physics to data science I m really happy I don't have to care about numerical precision on my computations. The probl

After moving from theoretical high energy physics to data science I m really happy I don't have to care about numerical precision on my computations. The problem is that numerical errors when solving partial differential equations not only propagate but increase in magnitude during the propagation. If you are not careful you will end up with a 100% wrong answer at the end of a big computation.