CODEWITHSUNDEEP
python
Rick
Rick

Reputation: 79

error message : "uncondensed distance matrix"

I have implemented hrp as described in the link below. I run with different inputs. The code produces reasonable results but I get the error message:

"main:75: ClusterWarning: scipy.cluster: The symmetric non-negative hollow observation matrix looks suspiciously like an uncondensed distance matrix"

Can anybody explain what it means?

link for code: https://github.com/lcamposgarrido/data-science_projects/blob/master/others/hierarchical_risk_parity/HRP.ipynb

Upvotes: 3

Views: 2566

Answers (1)

rmbrad
rmbrad

Reputation: 1102

The scipy.cluster.hierarchy.linkage function accepts either a 1-D condensed distance matrix or a 2-D array of observation vectors. The warning just means your passing a 2-D matrix that looks like a redundant distance matrix (non-negative, symmetric, diagonal all zeros), but it is being treated as a 2-D observation matrix. If it is actually a distance matrix it should be converted to the condensed form before passing to scipy.cluster.hierarchy.linkage. The scipy.spatial.distance.squareform function may be used to convert between the two distance matrix forms.

Upvotes: 1

Related Questions