Issues accessing class attributes of external package distance function

Dear Weka developers,

 

I am trying to fetch the result of a distance metric, implemented as Weka
external package, and use it in python via python-weka-wrapper3
<https://github.com/fracpete/python-weka-wrapper3> .

 

 

For this, I augment DilcaDistance.java by the method getMatricesDilca (and
build a jar myself), as shown below (most parts of the file are omitted):

 
<https://svn.cms.waikato.ac.nz/svn/weka/trunk/packages/external/DilcaDistanc
e/src/main/java/weka/core/DilcaDistance.java>
https://svn.cms.waikato.ac.nz/svn/weka/trunk/packages/external/DilcaDistance
/src/main/java/weka/core/DilcaDistance.java 

public class DilcaDistance implements DistanceFunction, Serializable{

 

    // ...

    

    protected Vector<double[][]> matricesDilca;

 

    //...

 

    public Object[] getMatricesDilca() {

        return matricesDilca.toArray();

    }

}

matricesDilca is computed for a given data set as core functionality of
DilcaDistance. I need to work with this set of matrices that represent
distances between categorical feature values, for all features in the data.

 

 

 

In the python context, I proceed as follows:

Employed versions:

*	python = "3.9.*"
*	python-weka-wrapper3 = "0.2.10"
*	python-javabridge = "4.0.3"

 

Find an excerpt of my Python code, containing a small example, below:

 

import numpy as np

import javabridge

import weka.core.jvm as jvm

from weka.filters import Filter

from weka.core.distances import DistanceFunction

from weka.core.typeconv import jdouble_matrix_to_ndarray

from weka.core.dataset import create_instances_from_matrices

 

 

class DilcaDistance(DistanceFunction):

    def __init__(self, supervised):

        _jobject = DistanceFunction.new_instance('weka.core.DilcaDistance')

        self.enforce_type(_jobject, 'weka.core.DistanceFunction')

        super().__init__(jobject=_jobject, options=None)

        self.is_optionhandler = True

        javabridge.call(self.jobject, 'setSupervisedDiscretization', '(Z)V',
supervised)

 

    def get_matrices_dilca(self):

        """Fetches dilca matrices from Java and loads those as numpy arrays.

        Matrix computation is done on a batch of data instances set via
self.instances.

 

        Returns:

            list: list of dilca matrices (numpy.ndarray each)

        """

        if not self.instances:

            raise ValueError('No instances provided.')

 

        jobj_arr = javabridge.call(self.jobject, 'getMatricesDilca',
'()[Ljava/lang/Object;')  # calling the new java method

        jmatrices = javabridge.get_env().get_object_array_elements(jobj_arr)

        matrices_dilca = []

       for jmat in jmatrices:

            matrices_dilca.append(jdouble_matrix_to_ndarray(jmat))

        return matrices_dilca

 

 

if not jvm.started:

    jvm.start(packages=True)

 

data = np.random.randint(1, 10, size=(10,5))

# print(data)

num2nom = Filter('weka.filters.unsupervised.attribute.NumericToNominal',

                 options=['-R', '1-5'])

ds_weka = create_instances_from_matrices(data, name='data')

num2nom.inputformat(ds_weka)

ds_weka = num2nom.filter(ds_weka)  # interpreting generated data as nominal

# print(ds_weka)

 

DD = DilcaDistance(supervised=False)

DD.instances = ds_weka

dm = DD.get_matrices_dilca()

print(dm)

 

jvm.stop()

 

 

I feel, there is a need to access the set of matricesDilca somehow. In my
case, this needs to occur from Python in order to integrate with other
elements. Is there a more elegant solution (maybe even without having to
augment the java code – please forgive my limited experience with Java)?

If not, is there any way to include the above augmentation in java (method
getMatricesDilca) in the official/public code? The reason for this is that,
alongside a publication of mine, I would like to reference the Weka
implementation for the sake of reproducibility of my findings.

 

Best

Martin

…………………………………………………

Martin Trat, M.Sc.

Wissenschaftlicher Mitarbeiter | Research Associate

Intelligent Systems and Production Engineering (ISPE)

 

FZI Forschungszentrum Informatik

Haid-und-Neu-Str. 10–14

76131 Karlsruhe, Germany

Tel.: +49 721 9654-509

[email protected]

 

 <https://www.fzi.de/team/martin-trat> www.fzi.de/team/martin-trat 

 <https://www.fzi.de/> www.fzi.de |  <http://www.twitter.com/FZI_official>
www.twitter.com/FZI_official |
<http://www.linkedin.com/company/fzi-official>
www.linkedin.com/company/fzi-official |
<https://www.youtube.com/FZIchannel> www.youtube.com/FZIchannel

…………………………………………………
FZI Forschungszentrum Informatik
Stiftung des bürgerlichen Rechts
Stiftung Az: 14-0563.1 Regierungspräsidium Karlsruhe
Vorstand: Prof. Dr. Andreas Oberweis, Jan Wiesenberger, Prof. Dr.-Ing. J.
Marius Zöllner
Vorsitzender des Kuratoriums: Ministerialdirigent Günther Leßnerkraus
…………………………………………………

_______________________________________________
Wekalist mailing list -- [email protected]
Send posts to [email protected]
To unsubscribe send an email to [email protected]
To subscribe, unsubscribe, etc., visit https://list.waikato.ac.nz/postorius/lists/wekalist.list.waikato.ac.nz
List etiquette: http://www.cs.waikato.ac.nz/~ml/weka/mailinglist_etiquette.html