Dear R-ians, I have a datamatrix X of nrows (large) and 3 colums. The 3 columns represent the coordinates of a cloud in a 3 dimensional space. Now I want to find the extremes of my cloud by using the N-FINDR algoritm. This algorithm consists of maximizing the volume between tree points. Therefore I need an optimaztion script, but I don't have any idea how to program it. According to literature the maximization of the volume can be replaced by maximizinng the determinant of E, where: E<-matrix(c(1,e1,1,e2,1,e3), ncol=3) and e1 to e3 are initially column vectors of the coordinates of random points. The procedure works by replacing an existing e1, e2 or e2 with another trial point. If the replacement results in an increase of determinant(E), the newly chosen pixel will replace the previous pixel. The procedure has to be repeated till there are no possible replacements left. In this way I get the coordinates of my extremes by e1 to e3. Anyone any idea how to program this? Kind regards and thanx in advance! Stef