Speeding up spectral clustering in Alibaba Call Graphs

[kapilagrawal95]

Hi,

I am rerunning the Alibaba call graph on all tracefiles (145 files) and am bottlenecked at the spectral clustering step because it takes a long time to finish. Is there a way to speed it up? I did try to use parfor to compute the Calinski-Harbasz score for every cluster_number but even that seems to be very slow.

Any suggestion on this would be helpful!

Thanks

Edit: I found this paper (https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=8661522) and its implementation (https://codeocean.com/capsule/0341996/tree/v2) to speed it up.