Graphlet Sampling kernel errors when sampling is 'None'

[Alessi0X]

I am using the Graphlet Sampling kernel with no sampling, i.e.

GK = grakel.GraphletSampling(n_jobs=None, normalize=False, verbose=False, random_state=None, k=this_k, sampling=None)

where this_k is user-defined.

However, when I do GK.fit_transform() on a dataset, it shows the following error

Traceback (most recent call last):
  File "/opt/rh/rh-python36/root/usr/lib64/python3.6/multiprocessing/pool.py", line 119, in worker
    result = (True, func(*args, **kwds))
  File "/opt/rh/rh-python36/root/usr/lib64/python3.6/multiprocessing/pool.py", line 44, in mapstar
    return list(map(*args))
  File "/opt/rh/rh-python36/root/usr/lib64/python3.6/site-packages/scipy/optimize/_differentialevolution.py", line 1265, in __call__
    return self.f(x, *self.args)
  File "main_benchmarkKernels.py", line 68, in myFitness
    thisKernelMatrix = GK.fit_transform(thisDataset)
  File "/opt/rh/rh-python36/root/usr/lib64/python3.6/site-packages/grakel/kernels/graphlet_sampling.py", line 309, in fit_transform
    self.fit(X)
  File "/opt/rh/rh-python36/root/usr/lib64/python3.6/site-packages/grakel/kernels/kernel.py", line 117, in fit
    self.initialize()
  File "/opt/rh/rh-python36/root/usr/lib64/python3.6/site-packages/grakel/kernels/graphlet_sampling.py", line 229, in initialize
    self.n_samples_ = n_samples
UnboundLocalError: local variable 'n_samples' referenced before assignment

By looking at the code (graphlet_sampling.py) a variable n_samples is assigned to self.n_samples_ even when sampling is None (line 229).

Maybe it's just a minor if/else issue? The if branch at line 158 considers the case when sampling is None and ends at line 160. The elseif branch starts at 161 and indeed declares n_samples, which can be assigned to self.n_samples at line 229.

Setup: Python 3.6.9 with grakel-dev==0.1a6

[ysig]

Can you try installing the upcoming version of grakel 0.1a7 https://github.com/ysig/GraKeL/tree/0.1a7. I think this error was a typo when refactoring some code and it is fixed there.

Thanks!

Στις Σάβ, 1 Φεβ 2020 στις 9:00 μ.μ., ο/η Alessio notifications@github.com έγραψε:

I am using the Graphlet Sampling kernel with no sampling, i.e.

GK = grakel.GraphletSampling(n_jobs=None, normalize=False, verbose=False, random_state=None, k=this_k, sampling=None)

where this_k is user-defined.

However, it shows the following error

Traceback (most recent call last): File "/opt/rh/rh-python36/root/usr/lib64/python3.6/multiprocessing/pool.py", line 119, in worker result = (True, func(*args, *kwds)) File "/opt/rh/rh-python36/root/usr/lib64/python3.6/multiprocessing/pool.py", line 44, in mapstar return list(map(args)) File "/opt/rh/rh-python36/root/usr/lib64/python3.6/site-packages/scipy/optimize/_differentialevolution.py", line 1265, in call return self.f(x, *self.args) File "main_benchmarkKernels.py", line 68, in myFitness thisKernelMatrix = GK.fit_transform(thisDataset) File "/opt/rh/rh-python36/root/usr/lib64/python3.6/site-packages/grakel/kernels/graphlet_sampling.py", line 309, in fit_transform self.fit(X) File "/opt/rh/rh-python36/root/usr/lib64/python3.6/site-packages/grakel/kernels/kernel.py", line 117, in fit self.initialize() File "/opt/rh/rh-python36/root/usr/lib64/python3.6/site-packages/grakel/kernels/graphlet_sampling.py", line 229, in initialize self.nsamples = n_samples UnboundLocalError: local variable 'n_samples' referenced before assignment

By looking at the code a variable n_samples is assigned to self.nsamples even when sampling is None.

Maybe it's just a minor if/else issue?

— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/ysig/GraKeL/issues/25?email_source=notifications&email_token=AGY7H2MC3OQIVIXI3BWKELDRAXBGVA5CNFSM4KOUZ52KYY3PNVWWK3TUL52HS4DFUVEXG43VMWVGG33NNVSW45C7NFSM4IKLJFUQ, or unsubscribe https://github.com/notifications/unsubscribe-auth/AGY7H2J63IUWBSUXE7EGERDRAXBGVANCNFSM4KOUZ52A .

[Alessi0X]

Will it be on pip soon? I forgot to mention that I installed grakel-dev via pip.

Anyway, by doing a diff on graphlet_sampling.py between versions 0.1a7 and 0.1a6 there is only one mismatch line

<             check_is_fitted(self, ['phi_X'])
---
>             check_is_fitted(self, ['_phi_X'])

so I don't think version 0.1a7 sorts out this issue.

[ysig]

Yes it will but for now you can install it from git using pip, as:

pip install https://github.com/ysig/GraKeL/archive/0.1a7.zip

Στις Κυρ, 2 Φεβ 2020 στις 9:14 μ.μ., ο/η Alessio notifications@github.com έγραψε:

Will it be on pip soon? I forgot to mention that I installed grakel-dev via pip

— You are receiving this because you commented. Reply to this email directly, view it on GitHub https://github.com/ysig/GraKeL/issues/25?email_source=notifications&email_token=AGY7H2MCITV3CLHFKSKC5PDRA4LTDA5CNFSM4KOUZ52KYY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOEKR6PCA#issuecomment-581166984, or unsubscribe https://github.com/notifications/unsubscribe-auth/AGY7H2M7KYBGJOTPF5PMCFTRA4LTDANCNFSM4KOUZ52A .

[Alessi0X]

Hi ysig. I tried 0.1a7 and still the same exact error.

[ysig]

Hey @Alessi0X.

Firstly, sorry for answering the same thing twice. It was confusion.

Secondly, thanks a lot for the feedback. It is actually a bug from refactoring. I think it is fixed with my latest commit (please have a look :) and close this issue if so).

Thirdly, I do know which is your application, but generally not sampling graphlets, is considered a bad practice. It takes a lot of time and doesn't return good results.

[Alessi0X]

Hi @ysig , thanks for you prompt reply. We now have a different error

"""
Traceback (most recent call last):
  File "/usr/local/Cellar/python/3.7.6_1/Frameworks/Python.framework/Versions/3.7/lib/python3.7/multiprocessing/pool.py", line 121, in worker
    result = (True, func(*args, **kwds))
  File "/usr/local/Cellar/python/3.7.6_1/Frameworks/Python.framework/Versions/3.7/lib/python3.7/multiprocessing/pool.py", line 44, in mapstar
    return list(map(*args))
  File "/usr/local/lib/python3.7/site-packages/scipy/optimize/_differentialevolution.py", line 1265, in __call__
    return self.f(x, *self.args)
  File "<ipython-input-2-0c9b4930054e>", line 40, in myFitness
    thisKernelMatrix = GK.fit_transform(thisDataset)
  File "/usr/local/lib/python3.7/site-packages/grakel/kernels/graphlet_sampling.py", line 310, in fit_transform
    self.fit(X)
  File "/usr/local/lib/python3.7/site-packages/grakel/kernels/kernel.py", line 123, in fit
    self.X = self.parse_input(X)
  File "/usr/local/lib/python3.7/site-packages/grakel/kernels/graphlet_sampling.py", line 416, in parse_input
    samples = self.sample_graphlets_(A, self.k_, self.n_samples_, self.random_state_)
TypeError: sample_graphlets() takes 1 positional argument but 4 were given
"""

[ysig]

I think now it works (I made a new commit). Can you confirm?

[Alessi0X]

Yes indeed, works nicely.

Thanks!

Closing the issue