@INPROCEEDINGS{Wagner_04,
    AUTHOR      = {Alan Wagner and James Chilson and Raymond Ng and Ruben Zamar},
    TITLE       = {{Parallel Computation of High Dimensional Robust Correlation
                    and Covariance Matrices}},
    BOOKTITLE   = {{10th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining}},
    YEAR        = {2004},
    EDITOR      = {},
    PAGES       = {},
    PUBLISHER   = {},
    VOLUME      = {},
    NUMBER      = {},
    SERIES      = {},
    ADDRESS     = {Seattle, WA},
    MONTH       = {August},
    NOTE        = {},
    KEYWORDS    = {},
    ISBN        = {},
    URL         = {http://hajek.stat.ubc.ca/~ruben/website/co.pdf},
    ABSTRACT    = {We describe two parallel algorithms for computing the correlation and covariance matrix. The first
           algorithm uses the Quadrant Correlation method and the second uses the Maronna method. The Maronna
           method obtains more robust covariance estimates, however, it is seldom used in practice because of the
           amount of computation required. Our parallel implementation of these methods makes them applicable
           to much larger problems, and in particular high dimensional datasets. The task farming approach to
           parallelizing the Maronna method also makes it feasible for large problems. We experimental evaluate
           the performance of these algorithms and use them in a real-life application related to cardiovascular
           research involving a large gene dataset with 6068 variables.},
}