3 ms·
Are you sure rounding error builds up exponentially? I may be wrong but I thought it was sqrt(N) where N is the number of operations.
by dnautics 9y ago
Are you sure rounding error builds up exponentially? I may be wrong but I thought it was sqrt(N) where N is the number of operations.
- dragontamer 9y agoIt has been a long time since I've taken the class. But my understanding is that the absolute error builds up exponentially. We both might be still correct however: sqrt(N) sounds about the speed of the "number of bits of error". So it depends on how you're counting error. If you're counting "bits of error", then sqrt(N) sounds right. If you're counting absolute error, then I'm pretty sure its exponential.