Table Methods for Continuous Random Variates

Abstract

We have Illustrated how algorithms can be sped up if we are willing to compute certain constants beforehand. For example, when a discrete random varlate is generated by the inversion method, it pays to compute and store the individual probabilities p _n beforehand. This information can speed up sequential search, or could be used in the method of guide tables. For continuous random varlates, the same remains true. Because we know many ultra fast discrete random varlate generation methods, but very few fast continuous random varlate generation techniques, there is a more pressing need for acceleration in the continuous case. Globally speaking, dlscretlzlng the problem speeds generation.