[SciPy-user] Origin of limits

[Stephen Walton]

On Mon, 2003-10-06 at 12:28, Scott Ransom wrote:

> Actually, epsilon is not the smallest representable number, it is the 
> smallest (fractional) difference you can have between 2 consecutive 
> double precision numbers and still have them be distinct from one 
> another.

To be even more precise, machine epsilon is defined as the smallest
value which one can add to 1 in floating point and get an answer which
is larger than 1.  Kahaner, Moler and Nash, "Numerical Methods and
Software", Prentice Hall Press.  Probably also Numerical Recipes and
many other places.

-- 
Stephen Walton <stephen.walton at csun.edu>
Dept. of Physics & Astronomy, Cal State Northridge