DD_tools: ball_dd

BALL_DD L_p ball description

W = BALL_DD(X,FRACREJ,P)

Fit a L_p ball around the data X by optimizing the weights: min w_0 s.t. \sum_j w_j|x_ij-a_j|^p <= w_0 \sum_j w_j = 1, w_j>=0 The vector a is taken as the mean of dataset X.

When the (feature-) weigths w are optimized, the threshold w_0 is set such that FRACREJ of the objects are outside the L_p ball.

As a precaution, features with no variance will be removed/ignored, otherwise a trivial solution of only using that feature is found. You will get a warning though.

See also: lpdd svdd myproxm Copyright: D.M.J. Tax, D.M.J.Tax@prtools.org Faculty EWI, Delft University of Technology P.O. Box 5031, 2600 GA Delft, The Netherlands