Remez algorithm: Difference between revisions

Revision as of 20:51, 8 June 2019

Revision as of 22:39, 9 June 2019

Line 106:

Sometimes more than one sample point is replaced at the same time with the locations of nearby maximum absolute differences.

Sometimes all of the sample points are replaced in a single iteration with the locations of all, alternating sign, maximum differences.<ref name=toobs>2/73 “The Optimization of Bandlimited Sytems”Systems” – G. C. Temes, F. C. Marshall and V. Barcilon. Proceedings IEEE.</ref>

Sometimes [[relative error]] is used to measure the difference between the approximation and the function, especially if the approximation will be used to compute the function on a computer which uses [[floating point]] arithmetic.

Sometimes zero-error point contraints are included in a Modified Remez Exchange Algorithm.<ref>2 name=toobs/73 “The Optimization of Bandlimited Sytems” – with G. C. Temes, F. C. Marshall, and V. Barcilon. Proceedings IEEE.<ref>

==See also==