Hi all, I would like to contribute to Mahout by starting with a simple task.
I had the idea of Implementing the Mahalanobis distance : http://en.wikipedia.org/wiki/Mahalanobis_distance This should be quite easy and could be a useful feature. After looking at the Mahout code for a couple of hours, I have three questions: 1) Generally speaking, what is the level of reliability of algorithms implemented in matrix.linalg ? 2) As inverting a covariance matrix is required, is using the method inverse(DoubleMatrix2D A) from class Algebra a good way to achieve this ? 3) Does is look reasonable if the computation of the distance is based on dense matrices ( inside the method distance(Vector v1,Vector v2) method part of the future MahalanobisDistanceMeasure class) ? 4) What is the best way to create a DoubleMatrix1D from a Vector ? Thanks a lot for your help. Cheers, Nicolas
