We want to find the covariance matrix with minimum volume encompassing some % of the data set. Let us assume we have a set of observations X = {x1, x2, … xN} of size N. Further let us assume 10% of samples are outliers. One could construct a brute-force algorithm as follows: