Added benchmarking script for Hierarchical CLustering, test script fo…

[chogba]

…r checking the benchmarking script, made corresponding additions to config.yaml and tests/tests.py

[mlpack-jenkins]

Can one of the admins verify this patch?

[chogba]

Have added a benchmarking script for shogun's hierarchical clustering implementation along with a testing script to test the benchmarking script

[karlnapf]

@chogba is there any other lib's hierarchical clustering algorithm present? I don't think so. I mean this is useful to get in, but getting algorithms in which then can be compared to other libs would be even more useful :) Ill make a few comments, leaving the rest to the mlpack crew

[chogba]

@karlnapf yep, makes sense, i'm working on implementing some other shogun models which other libraries have also implemented

[zoq]

@mlpack-jenkins test this please

[zoq]

hm, currently we test against shogun 6.0.0 (https://github.com/mlpack/benchmarks/blob/master/libraries/package-urls.txt), maybe hierarchical clustering was introduced in > 6.0.0?

[chogba]

It says "no module named hierarchical_clustering" I feel it's some bad naming on my part Let me just check once again and get back with the suggestions incorporated as well

[chogba]

I have no clue as to why the MATLAB tests are failing though Didn't change anything there

[rcurtin]

Sorry, that was my fault, I rebooted the benchmarking test systems and did not make sure the MATLAB tunnels came back up correctly. I'll try the build again...

[rcurtin]

@mlpack-jenkins test this please

[chogba]

I had run tests.py locally and it hadn't thrown any error I can't figure out why it can't find the module Could you once check if there are any path/naming issues? I'm on it as well

[rcurtin]

The MATLAB tests are now working.

[zoq]

@chogba I'm not sure this is a path issue, since every other shogun test runs fine; do you also test with shogun 6.0.0?

[rcurtin]

I left a few comments, let me know what you think. Happy to add more things to benchmark, so thank you for the contribution. :)

[chogba]

@zoq the src folder of shogun 6.0.0 shows a Hierarchical.cpp+.h combination so I think Hierarchical clustering is a part of the module @karlnapf is that correct?

[zoq]

Okay, I think if you adjust the path as I pointed out above the test should run just fine.

[chogba]

@zoq @rcurtin kindly review now

[zoq]

@mlpack-jenkins test this

[chogba]

my bad again

[chogba]

should be fine now, sorry for the trouble

[zoq]

@mlpack-jenkins add to whitelist

[chogba]

Was there still a problem?

[chogba]

@zoq @rcurtin kindly let me know if there's an issue

[chogba]

@zoq @rcurtin any updates?

[chogba]

@karlnapf @zoq @rcurtin kindly update, so that i can proceed with your inputs, and contribute further... thank you!

[zoq]

Hello @chogba can you also test the other distances: manhattan, cosine, chebyshev. Also, it's fine for me to merge this afterwards, we can add additional metrics once we figured out what makes sense in this case. How does this sound?

[chogba]

@zoq works fine! will add the tests for other distance checks too.. Also,I am working on adding a similar benchmark script for another algorithm in the meantime.. Could you provide me with a few more pointers as to where else I can devote my time?

[rcurtin]

@chogba don't worry, I see every update to every PR in an email. I respond when I have time (there are lots of PRs), so please be patient.

If there are no obvious other metrics, I agree with @zoq that we can wait until later to add them, if necessary, but let's see if @karlnapf has any input on the matter.

Right now the continuous integration for this repository is offline (hopefully it will be back up today or tomorrow); when it is back online, I'll have Jenkins test this and then assuming that all works I think there will be no more comments from my end.

[karlnapf]

lgtm!

[rcurtin]

@mlpack-jenkins test this please

[karlnapf]

Same here! Thx guys!