In the paper Outlier Detection for High Dimensional Data at the beginning of section 1.3 Is written:

Each attribute of the data is divided into $\phi$ equi-depth ranges. Thus, each range contains a fraction f = 1 /$\phi$ of the records.

What does it mean? Can you write an example?
If the data is $n \times p$ will I get $\phi$ matrix of size $\dfrac{n}{\phi}\times p$?

  • $\begingroup$ It seems to me the second sentence of the quotation includes the answer to the question. $\endgroup$ – whuber Aug 20 '14 at 13:30
  • $\begingroup$ After a while my understanding is that each variables have to be split in $\phi$ parts independently from the others. $\endgroup$ – Donbeo Aug 20 '14 at 13:54

Consider a matrix where the rows represent entries and columns attributes. An equi-depth split would be a split of phi boxes, where each box would contain some elements with all their attributes.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.