0

Hello,

Outlier detection is a huge problem with many applications and specifities. But let us be general about that issue and let's consider a metric space. What is in your opinion the best definition of an outlier in this case ?

Guillaume

flag
3 
 Mathoverflow is (mostly) about asking specific questions that have definitive answers, please read the faq. – Benoît Kloeckner Mar 9 2011 at 8:54
1 
 The question as stated is very vague; what exactly is the goal? Are "outliers" merely psychologically outliers, or they are points very "far" from the center of gravity, and so on. – S. Sra Mar 9 2011 at 8:57
the goal is to understand how people define an outlier, what the general idea is behind the technics they use. For example, a very simple definition would be : $\text{a point p is an outlier of a dataset D if its distance from the center of gravity of D is above a threshold \lambda}$. But maybe, someone has a better, more complex definition If mathoverflow is not the good place to ask this question, then where should i ask this ? – GuillaumeThomas Mar 9 2011 at 9:56
You are probably better off asking this question at the statistics StackExchange website: stats.stackexchange.com – Willie Wong Mar 9 2011 at 12:26

closed as subjective and argumentative by Benoît Kloeckner, Bill Johnson, Gjergji Zaimi, Mark Sapir, Willie Wong Mar 9 2011 at 12:25

Browse other questions tagged or ask your own question.