[MATH-948] Add inverse error function and inverse complementary error function

• Type: Wish
• Status: Closed
• Priority: Minor
• Resolution: Fixed
• Affects Version/s: 3.1.1
• Fix Version/s: 3.2
• Assignee: Luc Maisonobe
• Reporter: Luc Maisonobe
• Created: 15/Mar/13 11:28
• Updated: 07/Apr/13 09:20
• Resolved: 15/Mar/13 11:42

Description

The Erf class provides only direct functions erf(double) and erfc(double).

The inverse of these functions are useful to compute accurately and efficiently the inverse cumulative probabilities of normal distribution and Lévy distribution.

A recent implementation as been designed by Mike Giles and published in a 2010 paper in CUDA Gems: Approximating the erfinv function. The paper is available at http://people.maths.ox.ac.uk/gilesm/files/gems_erfinv.pdf. The code implementing the double precision version with all the coefficients is available here: http://gpucomputing.net/?q=node/1828.

After a direct contact with professor Giles, he gave his authorization to the Apache Commons project to use this code in the library.