[scikit-learn] Smoke and Metamorphic Testing of scikit-learn
herbold at cs.uni-goettingen.de
Mon Aug 27 09:07:03 EDT 2018
I now have results for LinearDiscriminantAnalysis and the SGDClassifier.
I updated the results online.
The LinearDiscriminantAnalysis had
* an infinity of NaN for data that approaches MAXDOUBLE and
* problems with an internal array size computation for data for
several tests, i.e., data that is very close to zero and cannot be
expressed by 32bit floats as well as for data that is all zero.
The SGD had
* an over/underflow for data that approaches MAXDOUBLE
* differences in the classifications if we added one to the numeric
* differences in the classification if we reordered the instances.
Am 23.08.2018 um 13:39 schrieb Steffen Herbold:
> Hi Andy,
> thanks for your detailed feedback.
> The random states are fixed, and set immediately before calling the
> fit function. Here is a gist with the code for one smoke tests and a
> metamorphic test .
> I will run the tests for LinearDiscriminantAnalysis and the
> SGDClassifier. I somehow missed them when I scanned the documentation.
> I know that these problems should sometimes be expected. However, I
> was actually not sure what to expect, especially after I started to
> look at the results for different ML libraries in comparison. The
> random forest you brought up are good example. I also expected them to
> be dependent on feature/instance order. However, they are not in Weka,
> only in scikit-learn and Spark MLlib. There are more such examples,
> like logistic regression that exihibits different behavior in all
> three libraries.
> I already have a comparison regarding expected differences between
> machine learning frameworks planned as a topic for future work.
>  https://gist.github.com/sherbold/570c9399e9bc39dd980d6c2bdbf3b64a
> Am 22.08.2018 um 17:49 schrieb Andreas Mueller:
>> Hi Steffen.
>> Thanks for sharing your analysis. We really need more work in this
>> I assume you fixed the random states everywhere?
>> I consider these tests helpful but not all your expectations are
>> warranted depending on the model.
>> If you add one to each feature, there is no expectations that results
>> will be the same, unless for the tree models.
>> For tree-based models with fixed random states, however, it's
>> expected that reordering features will change the result.
>> For non-convex optimization it's expected that results are not
>> symmetric (i.e. the MLPClassifier will not flip
>> the decision function because the optimization is initialized in an
>> asymetric way), and reordering features will
>> also change the result. If using mini-batches (the default) the
>> results will also change when instances are reordered.
>> I assume you didn't test SGDClassifier or any of it's derivatives
>> because it doesn't show up here. Did you test
>> For the invariance tests it would be interesting to know if they are
>> due to tie-breaking or numerical issues.
>> There is some numerical issues that are very hard to control, and I'm
>> pretty sure we have asymmetric tie-breaking
>> (multiclass libsvm is "always predict the first class"
>> https://github.com/scikit-learn/scikit-learn/issues/8276 )
>> I would looks at QuadraticDiscriminantAnalysis a bit more closely as
>> a consequence of your tests.
>> Maybe check if the SVM, RF and KNN issues are due to tie-breaking.
>> We could try and document all the cases where the result will not
>> fulfill these invariances, but I think that might be too much.
>> At some point we need the users to understand what's going on. If you
>> look at the random forest algorithm and you fix
>> the random state it's obvious that feature order matters.
>> A big question here is how big the differences are. Some algorithms
>> are randomized (I think the coordinate descent in
>> some of the linear models uses random orders), but the results are
>> expected to be near-identical, independent of the ordering.
>> On 8/22/18 7:12 AM, Steffen Herbold wrote:
>>> Dear developers,
>>> I am writing you because I applied an approach for the automated
>>> testing of classification algorithms to scikit-learn and would like
>>> to forward the results to you.
>>> The approach is a combination of smoke testing and metamorphic
>>> testing. The smoke tests try to find problems by executing the
>>> training and prediction functions of classifiers with different
>>> data. These smoke tests should ensure the basic functioning of
>>> classifiers. I defined 20 different data sets, some very simple
>>> (uniform features in [0,1]), some with extreme distributions, e.g.,
>>> data close to machine precision. The metamorphic tests determine if
>>> classification results change as expected if the training data is
>>> modified, e.g., by reordering features, flipping class labels, or
>>> reordering instances.
>>> I generated 70 different Python unittest tests for eleven different
>>> scikit-learn classifiers. In summary, I found the following
>>> potential problems:
>>> - Two errors due to possibly infinite loops for the
>>> LogisticRegressionClassifier for data that approaches MAXDOUBLE.
>>> - The classification of LogisticRegression, MLPClassifier,
>>> QuadraticDiscriminantAnalysis, and SVM with a polynomial kernel
>>> changed if one is added to each feature value.
>>> - The classification of DecisionTreeClassifier, LogisticRegression,
>>> MLPClassifier, QuadraticDiscriminantAnalysis,
>>> RandomForestClassifier, and SVM with a linear and a polynomial
>>> kernel were not inverted when all binary class labels are flipped.
>>> - The classification of LogisticRegression, MLPClassifier,
>>> QuadraticDiscriminantAnalysis, and RandomForestClassifier sometimes
>>> changed when the features are reordered.
>>> - The classification of KNeighborsClassifier, MLPClassifier,
>>> QuadraticDiscriminantAnalysis, RandomForestClassifier, and SVM with
>>> a linear kernel sometimes changed when the instances are reordered.
>>> You can find details of our results online . The provided
>>> resources include the current draft of the paper that describes the
>>> tests as well as detailed results in detail. Moreover, we provide an
>>> executable test suite with all tests we executed, as well as the
>>> export of our test results as XML file that contains all details of
>>> the test execution, including stack traces in case of exceptions.
>>> The preprint and online materials also contain the results for two
>>> other machine learning libraries, i.e., Weka and Spark MLlib.
>>> Additionally, you can find the atoml tool used to generate the tests
>>> on GitHub .
>>> I hope that these tests may help with the future development of
>>> scikit-learn. You could help me a lot by answering the following
>>> - Do you consider the tests helpful?
>>> - Do you consider any source code or documentation changes due to
>>> our findings?
>>> - Would you be interested in a pull request or any other type of
>>> integration of (a subset of) the tests into your project?
>>> - Would you be interested in more such tests, e.g., for the
>>> consideration of hyper parameters, other algorithm types like
>>> clustering, or more complex algorithm specific metamorphic tests?
>>> I am looking forward to your feedback.
>>> Best regards,
>>> Steffen Herbold
>>>  http://user.informatik.uni-goettingen.de/~sherbold/atoml-results/
>>>  https://github.com/sherbold/atoml
>> scikit-learn mailing list
>> scikit-learn at python.org
Dr. Steffen Herbold
Institute of Computer Science
University of Goettingen
37077 Göttingen, Germany
mailto. herbold at cs.uni-goettingen.de
tel. +49 551 39-172037
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the scikit-learn