What is the diference between Bregman Divergence Mahalanobis Distance and Mahalanobis Distance?

agucaba123

Hi, I'm a beginner in RapidMiner Studio. I'm working with clustering operators (k-means and k-medoid) and I want to know what is the diference between the Mahalanobis Distance (https://en.wikipedia.org/wiki/Mahalanobis_distance) and the option in the parameters of the operator which says "Bregman divergences" and the sub-option divergence: "Mahalanobis distance". I don't know if they are the same or not. 

DocMusher

Hi,

Perhaps these slides might explain your question? 

Cheers

Sven

MartinLiebig

Hi @agucaba123,

 

here is what we do:

	@Override
	public double calculateDistance(double[] value1, double[] value2) {

		Matrix x = new Matrix(value1, value1.length);
		Matrix y = new Matrix(value2, value2.length);

		Matrix deltaxy = x.minus(y);

		// compute the mahalanobis distance
		return Math.sqrt(deltaxy.transpose().times(inverseCovariance).times(deltaxy).get(0, 0));
	}

Does this make sense?

 

BR,

Martin