I've been using Mahalanobis distance to look for outliers. This link: https://www.cs.princeton.edu/courses/archive/fall08/cos436/Duda/PR_Mahal/M_metric.htm says that feature scaling is addressed in the computation of the Mahalanobis distance. Am I interpreting this correctly?
1
-
$\begingroup$ Please paste in whatever context is necessary to understand & answer your question. We want this thread to remain valuable even if the link goes dead. $\endgroup$– gung - Reinstate MonicaCommented Apr 30, 2016 at 13:04
Add a comment
|
1 Answer
$\begingroup$
$\endgroup$
3
The Mahalanobis distance considers the variance of all the features while finding out the distance between two points, as is clear from the formula.
Feature scaling, generally, means that mean-centering and division by the standard deviation of the feature. If you scale all the feature and then find out the euclidean distance between two points, you will actually Mahalanobis distance.
-
$\begingroup$ But the "correct" feature scaling is the difficult point, right? $\endgroup$– BenCommented Nov 21, 2017 at 6:55
-
$\begingroup$ @Ben What do you mean by 'correct'? $\endgroup$ Commented Nov 22, 2017 at 5:41
-
$\begingroup$ Ah, I mixed up feature scaling and scaling. Thx! $\endgroup$– BenCommented Nov 22, 2017 at 7:41