search

ysig / GraKeL

A scikit-learn compatible library for graph kernels
https://ysig.github.io/GraKeL/
Other
588 stars 96 forks source link

Dealing with multiple sets of continuous attributes? #41

Closed SylvainTakerkart closed 4 years ago

SylvainTakerkart commented 4 years ago

Hi there!

First, nice job with GraKel ;) It took me very little time to get my graphs into a format that can be handled by GraKel (thanks to the networkx import method and the corresponding example)!

My graphs (computed from some brain imaging data) have multiple vector-valued attributes on each node... Can any of the kernels availabe in GraKel deal with this? (I of course have the solution of concatenating the different vectors into one, but I'm just wondering whether there's a cleaner solution...) Or any kernel in general?

Thanks,

Sylvain

giannisnik commented 4 years ago

Thank you @SylvainTakerkart . Unfortunately, there are no kernels that can handle graphs whose nodes are assigned multiple vector-valued attributes (2d matrices). As you mentioned, you can concatenate these vectors and then employ some kernel that takes them into account (e.g., GraphHopper, PropagationAttr, MultiscaleLaplacianFast, SubgraphMatching).

SylvainTakerkart commented 4 years ago

OK, thanks for your answer!