(thing) by wamckee Sat Apr 13 2002 at 17:17:35
{computer science} Epsilon is used to repesent the smallest value that a floating point number (a real number on the computer) can hold such that it satisfies the equation


    R (1.0 + epsilon) not equal to R (1.0)

    where the function R maps a real number to a
    representation that uses a finite amount of
    computer memory (limited precision).

It is often used as a stopping criteria for algorithms since a change less than epsilon would not result in the change in a floating point value as represented in the computer's memory.

Epsilon for an 8 byte IEEE floating point number is about 2.2 x 10-16 and is defined as DBL_EPSILON in the C standard library in the header file "float.h".

I like it!