Minimal variance algorithm. Ideas?

I don't have much experience with developing algorithms but I am considering the following problem: How to divide a set of numbers in N groups, such that the total variance is minimal? So the variance within each group is summed to get total variance.

Is it possible to calculate directly (like thresholding quantiles if I wanted groups of equal size) or does it require an iterative process?

Ive checked the code of rgb2ind but it uses some built-in functionality (cq?), and of course rgb2ind works on points in 3D space. So any help, even a reference to a handy paper or book, is welcome!

Answer 1

[Matt J]

http://en.wikipedia.org/wiki/K-means

http://www.mathworks.com/matlabcentral/fileexchange/index?utf8=%E2%9C%93&term=k-means