Computer Methods for Sampling from
the Exponential and Normal Distributions
Various methods are known for transforming uniformly
distributed random numbers into exponentially
and normally distributed quantities.  The most efficient
ones are compared, in terms of memory requirements 
and speed, with some new algorithms.  A number of procedures
convert Taylor series expansions directly 
into sampling steps, an approach which may be used for
sampling from any continuous distribution.  For 
the exponential distribution a definite recommendation
can be made, whereas in the case of the normal 
distribution there remains a choice between slower and
shorter algorithms and faster but space consuming 
methods.
CACM October, 1972
Ahrens, J. H.
Dieter, U.
