diff --git a/src/methods/softmax_regression/softmax_regression.cpp b/src/methods/softmax_regression/softmax_regression.cpp
index ddf2c2ce7134d7c30519d4031633188539336155..a4505f21a2b058a1881979db6c51b4b4fa8109be 100644
--- a/src/methods/softmax_regression/softmax_regression.cpp
+++ b/src/methods/softmax_regression/softmax_regression.cpp
@@ -19,6 +19,7 @@ using namespace mlpack::regression;
SoftmaxRegression softmaxRegression;
+
// input: const size_t inputSize = 0,
// const size_t numClasses = 0,
// const bool fitIntercept = false//
@@ -74,8 +75,7 @@ void initModelWithTrain(float *dataMatArr, SP_integer dataMatSize, SP_integer da
SP_integer classifyPoint(float *pointArr, SP_integer pointArrSize)
{
// convert the Prolog arrays to arma::rowvec
- rowvec pointVector = convertArrayToRowvec(pointArr, pointArrSize);
-
+ vec pointVector = conv_to<vec>::from(convertArrayToRowvec(pointArr, pointArrSize));
try
{
@@ -118,6 +118,17 @@ void classifyMatrix(float *dataMatArr, SP_integer dataMatSize, SP_integer dataMa
}
+ // check for nan elements
+ if (labelsReturnVector.has_nan())
+ {
+ raisePrologSystemExeption("Labels return Vector contains nan!");
+ return;
+ }
+ if (probsReturnMat.has_nan())
+ {
+ raisePrologSystemExeption("Probabilities return Matrix contains nan!");
+ return;
+ }
// return the Vector
returnVectorInformation(labelsReturnVector, labelsArr, labelsArrSize);
// return the Matrix
@@ -132,18 +143,24 @@ void classifyMatrix(float *dataMatArr, SP_integer dataMatSize, SP_integer dataMa
// description:
// Computes accuracy of the learned model given the feature data and the labels associated with each data point.
//
-double computeAccuracy(float *dataMatArr, SP_integer dataMatSize, SP_integer dataMatRowNum, float *labelsArr, SP_integer labelsArrSize)
+double computeAccuracy(float *dataMatArr, SP_integer dataMatSize, SP_integer dataMatRowNum,
+ float *labelsArr, SP_integer labelsArrSize)
{
// convert the Prolog array to arma::mat
mat data = convertArrayToMat(dataMatArr, dataMatSize, dataMatRowNum);
// convert the Prolog array to arma::rowvec
Row< size_t > labelsVector = convertArrayToVec(labelsArr, labelsArrSize);
-
+
try
{
return softmaxRegression.ComputeAccuracy(data, labelsVector);
}
+ catch(const std::out_of_range& e)
+ {
+ raisePrologSystemExeption("The Labels Vector has the wrong Dimension!");
+ return 0.0;
+ }
catch(const std::exception& e)
{
raisePrologSystemExeption(e.what());
@@ -188,7 +205,8 @@ void parameters(float **parametersMatArr, SP_integer *parametersMatColNum, SP_in
// description:
// Trains the softmax regression model with the given training data.
//
-double train(float *dataMatArr, SP_integer dataMatSize, SP_integer dataMatRowNum, float *labelsArr, SP_integer labelsArrSize, SP_integer numClasses)
+double train(float *dataMatArr, SP_integer dataMatSize, SP_integer dataMatRowNum,
+ float *labelsArr, SP_integer labelsArrSize, SP_integer numClasses)
{
// convert the Prolog array to arma::mat
mat data = convertArrayToMat(dataMatArr, dataMatSize, dataMatRowNum);
diff --git a/src/methods/softmax_regression/softmax_regression.pl b/src/methods/softmax_regression/softmax_regression.pl
index 4a34dd6b71d76ebf92836516bbb46655fca2c144..17e11624c49d642357995a3bf043b9c63483d944 100644
--- a/src/methods/softmax_regression/softmax_regression.pl
+++ b/src/methods/softmax_regression/softmax_regression.pl
@@ -25,7 +25,7 @@
%% --Input--
-%% int inputSize => 0,
+%% int inputSize => 1,
%% int numClasses => 0,
%% bool fitIntercept => (1)true / (0)false => false
%%
@@ -36,6 +36,8 @@
%% Be sure to use Train before calling Classif or ComputeAccuracy, otherwise the results may be meaningless.
%%
initModelNoTrain(InputSize, NumClasses, FitIntercept) :-
+ InputSize > 0,
+ NumClasses >= 0,
initModelNoTrainI(InputSize, NumClasses, FitIntercept).
foreign(initModelNoTrain, c, initModelNoTrainI( +integer, +integer,
@@ -55,6 +57,8 @@ foreign(initModelNoTrain, c, initModelNoTrainI( +integer, +integer,
%% Initializes the softmax_regression model and trains it.
%%
initModelWithTrain(DataList, DataRows, LabelsList, NumClasses, Lambda, FitIntercept) :-
+ NumClasses >= 0,
+ Lambda >= 0,
convert_list_to_float_array(DataList, DataRows, array(Xsize, Xrownum, X)),
convert_list_to_float_array(LabelsList, array(Ysize, Y)),
initModelWithTrainI(X, Xsize, Xrownum, Y, Ysize, NumClasses, Lambda, FitIntercept).
@@ -144,9 +148,9 @@ foreign(featureSize, c, featureSizeI([-integer])).
%% --Description--
%% Get the model parameters.
%%
-parameters(PraametersList, XCols) :-
+parameters(PrametersList, XCols) :-
parametersI(X, XCols, XRows),
- convert_float_array_to_2d_list(X, XCols, XRows, PraametersList).
+ convert_float_array_to_2d_list(X, XCols, XRows, PrametersList).
foreign(parameters, c, parametersI(-pointer(float_array), -integer, -integer)).
@@ -163,6 +167,7 @@ foreign(parameters, c, parametersI(-pointer(float_array), -integer, -integer)).
%% Trains the softmax regression model with the given training data.
%%
train(DataList, DataRows, LabelsList, NumClasses, FinalValue) :-
+ NumClasses >= 0,
convert_list_to_float_array(DataList, DataRows, array(Xsize, Xrownum, X)),
convert_list_to_float_array(LabelsList, array(Ysize, Y)),
trainI(X, Xsize, Xrownum, Y, Ysize, NumClasses, FinalValue).
diff --git a/src/methods/softmax_regression/softmax_regression_test.pl b/src/methods/softmax_regression/softmax_regression_test.pl
index 2ed582c0e0e8d404f64e03c990cddb4598adc118..6612ce1f49f7151a2bf2ba6af199bb31f27ee5e9 100644
--- a/src/methods/softmax_regression/softmax_regression_test.pl
+++ b/src/methods/softmax_regression/softmax_regression_test.pl
@@ -6,38 +6,335 @@
:- use_module(softmax_regression).
:- use_module('../../helper_files/helper.pl').
-reset_Model :-
- initModel(1,0,50,0.0001).
+reset_Model_NoTrain :-
+ initModelNoTrain(3, 2, 0).
+
+reset_Model_WithTrain :-
+ initModelWithTrain([5.1,3.5,1.4,4.9,3.0,1.4,4.7,3.2,1.3,4.6,3.1,1.5], 3, [0,1,0,1], 2, 0.0001, 0).
+
+%%
+%% TESTING predicate initModelNoTrain/3
+%%
+:- begin_tests(initModelNoTrain).
+
+%% Failure Tests
+
+test(softmax_InitNoTrain_Negative_InputSize, fail) :-
+ initModelNoTrain(-1, 0, 0).
+
+test(softmax_InitNoTrain_Negative_InputSize, fail) :-
+ initModelNoTrain(3, -1, 0).
+
+
+%% Successful Tests
+
+test(softmax_InitNoTrain_FitIntercept_False) :-
+ initModelNoTrain(3, 2, 0).
+
+test(softmax_InitNoTrain_FitIntercept_True) :-
+ initModelNoTrain(2, 3, 1).
+
+:- end_tests(initModelNoTrain).
+
%%
-%% TESTING predicate predicate/10
+%% TESTING predicate initModelWithTrain/6
%%
-:- begin_tests(predicate).
+:- begin_tests(initModelWithTrain).
%% Failure Tests
-test(testDescription, [error(domain_error('expectation' , culprit), _)]) :-
- reset_Model_No_Train(perceptron),
- train([5.1,3.5,1.4,4.9,3.0,1.4,4.7,3.2,1.3,4.6,3.1,1.5], 3, [0,0,0,0], 2, culprit, 50, 0.0001, _).
-test(testDescription2, [error(_,system_error('The values of the Label have to start at 0 and be >= 0 and < the given numClass!'))]) :-
- reset_Model_No_Train(perceptron),
- train([5.1,3.5,1.4,4.9,3.0,1.4,4.7,3.2,1.3,4.6,3.1,1.5], 3, [0,1,0,2], 2, perceptron, 50, 0.0001, _).
+test(softmax_InitWithTrain_Negative_NumClass, fail) :-
+ initModelWithTrain([5.1,3.5,1.4,4.9,3.0,1.4,4.7,3.2,1.3,4.6,3.1,1.5], 3, [0,1,0,1], -1, 0.0001, 0).
+
+test(softmax_InitWithTrain_Negative_Lambda, fail) :-
+ initModelWithTrain([5.1,3.5,1.4,4.9,3.0,1.4,4.7,3.2,1.3,4.6,3.1,1.5], 3, [0,1,0,1], 2, -0.0001, 0).
+
+test(softmax_InitWithTrain_Wrong_Label_Dims1, [error(_,system_error('element-wise multiplication: incompatible matrix dimensions: 2x4 and 2x2'))]) :-
+ initModelWithTrain([5.1,3.5,1.4,4.9,3.0,1.4,4.7,3.2,1.3,4.6,3.1,1.5], 3, [0,1], 2, 0.0001, 0).
+
+
+test(softmax_InitWithTrain_Wrong_Label_Dims2, [error(_,system_error('element-wise multiplication: incompatible matrix dimensions: 2x4 and 2x7'))]) :-
+ initModelWithTrain([5.1,3.5,1.4,4.9,3.0,1.4,4.7,3.2,1.3,4.6,3.1,1.5], 3, [0,1,0,1,0,0,1], 2, 0.0001, 0).
+
+%% Doesnt cause exception
+test(softmax_InitWithTrain_Wrong_Label_Value) :-
+ initModelWithTrain([5.1,3.5,1.4,4.9,3.0,1.4,4.7,3.2,1.3,4.6,3.1,1.5], 3, [0,-1,0,-1], 2, 0.0001, 0).
+
+%% Doesnt cause exception
+test(softmax_InitWithTrain_Too_Many_Label_Value) :-
+ initModelWithTrain([5.1,3.5,1.4,4.9,3.0,1.4,4.7,3.2,1.3,4.6,3.1,1.5], 3, [1,1,0,2], 2, 0.0001, 0).
%% Successful Tests
-test(testDescription3, [true(Error =:= 1)]) :-
- reset_Model_No_Train(perceptron),
- train([5.1,3.5,1.4,4.9,3.0,1.4,4.7,3.2,1.3,4.6,3.1,1.5], 3, [0,0,0,0], 2, perceptron, 50, 0.0001, Error).
+test(softmax_InitWithTrain_Direct_Input) :-
+ initModelWithTrain([5.1,3.5,1.4,4.9,3.0,1.4,4.7,3.2,1.3,4.6,3.1,1.5], 3, [0,1,0,1], 2, 0.0001, 0).
-test(testDescription4, [true(Error =:= 0.9797958971132711)]) :-
- reset_Model_No_Train(perceptron),
- open('/home/afkjakhes/eclipse-workspace/prolog-mlpack-libary/src/data_csv/iris2.csv', read, File),
+test(softmax_InitWithTrain_CSV_Input) :-
+ open('src/data_csv/iris2.csv', read, File),
take_csv_row(File, skipFirstRow,10, Data),
- train(Data, 4, [0,1,0,1,1,0,1,1,1,0], 2, perceptron, 50, 0.0001, Error).
+ initModelWithTrain(Data, 4, [0,1,0,1,1,0,1,1,1,0], 2, 0.003, 1).
+
+:- end_tests(initModelWithTrain).
+
+
+%%
+%% TESTING predicate classifyPoint/2
+%%
+:- begin_tests(classifyPoint).
+
+%% Failure Tests
+
+%% Doesnt cause an exception
+test(softmax_ClassifyPoint_On_Untrained_Model) :-
+ reset_Model_NoTrain,
+ classifyPoint([5.1,3.5,1.4], _).
+
+test(softmax_ClassifyPoint_With_Too_Big_Dims, [error(_,system_error('SoftmaxRegression::Classify(): dataset has 5 dimensions, but model has 3 dimensions!'))]) :-
+ reset_Model_WithTrain,
+ classifyPoint([5.1,3.5,1.4,5.2,3.2], _).
+
+test(softmax_ClassifyPoint_With_Too_Small_Dims, [error(_,system_error('SoftmaxRegression::Classify(): dataset has 2 dimensions, but model has 3 dimensions!'))]) :-
+ reset_Model_WithTrain,
+ classifyPoint([5.1,3.5], _).
+
+
+%% Successful Tests
+
+test(softmax_ClassifyPoint) :-
+ reset_Model_WithTrain,
+ classifyPoint([4.1,2.5,1.4], Prediction),
+ print('\nPrediction: '),
+ print(Prediction).
+
+:- end_tests(classifyPoint).
+
+
+%%
+%% TESTING predicate classifyMatrix/5
+%%
+:- begin_tests(classifyMatrix).
+
+%% Failure Tests
+
+%% Doesnt cause an exception
+test(softmax_ClassifyMatrix_On_Untrained_Model) :-
+ reset_Model_NoTrain,
+ classifyMatrix([3, 2, 0, 5, 1, 4, 1, 0, 4, 3, 3, 5, 0, 5, 5], 3, _, _, _).
+
+test(softmax_ClassifyMatrix_With_Too_Big_Dims, [error(_,system_error('SoftmaxRegression::Classify(): dataset has 5 dimensions, but model has 3 dimensions!'))]) :-
+ reset_Model_WithTrain,
+ classifyMatrix([3, 2, 0, 5, 1, 4, 1, 0, 4, 3, 3, 5, 0, 5, 5], 5, _, _, _).
+
+test(softmax_ClassifyMatrix_With_Too_Small_Dims, [error(_,system_error('SoftmaxRegression::Classify(): dataset has 2 dimensions, but model has 3 dimensions!'))]) :-
+ reset_Model_WithTrain,
+ classifyMatrix([3, 2, 0, 5, 1, 4, 1, 0, 4, 3, 3, 5, 0, 5], 2, _, _, _).
+
+
+%% Successful Tests
+
+test(softmax_ClassifyMatrix_Direct_Input) :-
+ reset_Model_WithTrain,
+ classifyMatrix([3, 2, 0, 5, 1, 4, 1, 0, 4, 3, 3, 5, 0, 5, 5], 3, PredicList, ProbsList, _),
+ print('\nPredicted Labels: '),
+ print(PredicList),
+ print('\nProbabilities: '),
+ print(ProbsList).
+
+test(softmax_ClassifyMatrix_CSV_Input) :-
+ open('src/data_csv/iris2.csv', read, File),
+ take_csv_row(File, skipFirstRow,10, Data),
+ initModelWithTrain(Data, 4, [0,1,0,1,1,0,1,1,1,0], 2, 0.003, 1),
+ classifyMatrix(Data, 4, PredicList, ProbsList, _),
+ print('\nPredicted Labels: '),
+ print(PredicList),
+ print('\nProbabilities: '),
+ print(ProbsList).
+
+:- end_tests(classifyMatrix).
+
+
+%%
+%% TESTING predicate computeAccuracy/4
+%%
+:- begin_tests(computeAccuracy).
+
+%% Failure Tests
+
+%% Doesnt cause an exception
+test(softmax_ComputeAccuracy_On_Untrained_Model) :-
+ reset_Model_NoTrain,
+ computeAccuracy([5.1,3.5,1.4,4.9,3.0,1.4,4.7,3.2,1.3,4.6,3.1,1.5], 3, [0,1,0,1], _).
+
+test(softmax_ComputeAccuracy_Wrong_Label_Dims1, [error(_,system_error('The Labels Vector has the wrong Dimension!'))]) :-
+ reset_Model_WithTrain,
+ computeAccuracy([5.1,3.5,1.4,4.9,3.0,1.4,4.7,3.2,1.3,4.6,3.1,1.5], 3, [0,1], _).
+
+%% Doesnt cause exception
+test(softmax_ComputeAccuracy_Wrong_Label_Dims2) :-
+ reset_Model_WithTrain,
+ computeAccuracy([5.1,3.5,1.4,4.9,3.0,1.4,4.7,3.2,1.3,4.6,3.1,1.5], 3, [0,1,0,1,0,0,1], _).
+
+%% The same when the label values are out of range
+test(softmax_ComputeAccuracy_Wrong_Label_Value) :-
+ reset_Model_WithTrain,
+ computeAccuracy([5.1,3.5,1.4,4.9,3.0,1.4,4.7,3.2,1.3,4.6,3.1,1.5], 3, [0,-1,0,-1], _).
+
+%% Doesnt cause an exception
+test(softmax_ComputeAccuracy_Too_Many_Label_Value) :-
+ reset_Model_WithTrain,
+ computeAccuracy([5.1,3.5,1.4,4.9,3.0,1.4,4.7,3.2,1.3,4.6,3.1,1.5], 3, [1,1,0,2], _).
+
+test(softmax_ComputeAccuracy_Wrong_Data_Dims, [error(_,system_error('SoftmaxRegression::Classify(): dataset has 4 dimensions, but model has 3 dimensions!'))]) :-
+ reset_Model_WithTrain,
+ computeAccuracy([5.1,3.5,1.4,4.9,3.0,1.4,4.7,3.2,1.3,4.6,3.1,1.5], 4, [0,1,0], Accuracy),
+ print('\nAccuracy: '),
+ print(Accuracy).
+
+
+%% Successful Tests
+
+test(softmax_ComputeAccuracy_Direct_Input) :-
+ reset_Model_WithTrain,
+ computeAccuracy([5.1,3.5,1.4,4.9,3.0,1.4,4.7,3.2,1.3,4.6,3.1,1.5], 3, [0,1,0,1], Accuracy),
+ print('\nAccuracy: '),
+ print(Accuracy).
+
+test(softmax_ComputeAccuracy_CSV_Input) :-
+ open('src/data_csv/iris2.csv', read, File),
+ take_csv_row(File, skipFirstRow,10, Data),
+ initModelWithTrain(Data, 4, [0,1,0,1,1,0,1,1,1,0], 2, 0.003, 1),
+ computeAccuracy([3, 2, 0, 5, 1, 4, 1, 0, 4, 3, 3, 5, 0, 5, 5, 2], 4, [0,1,0,1], Accuracy),
+ print('\nAccuracy: '),
+ print(Accuracy).
+
+:- end_tests(computeAccuracy).
+
+
+%%
+%% TESTING predicate featureSize/1
+%%
+:- begin_tests(featureSize).
+
+%% Failure Tests
+
+
+
+
+%% Successful Tests
+
+test(softmax_FeatureSize_No_Train) :-
+ reset_Model_NoTrain,
+ featureSize(FeatureSize),
+ print('\nFeatureSize: '),
+ print(FeatureSize).
+
+test(softmax_FeatureSize_Direct_Input) :-
+ reset_Model_WithTrain,
+ featureSize(FeatureSize),
+ print('\nFeatureSize: '),
+ print(FeatureSize).
+
+test(softmax_FeatureSize_CSV_Input) :-
+ open('src/data_csv/iris2.csv', read, File),
+ take_csv_row(File, skipFirstRow,10, Data),
+ initModelWithTrain(Data, 4, [0,1,0,1,1,0,1,1,1,0], 2, 0.003, 1),
+ featureSize(FeatureSize),
+ print('\nFeatureSize: '),
+ print(FeatureSize).
+
+:- end_tests(featureSize).
+
+
+%%
+%% TESTING predicate parameters/2
+%%
+:- begin_tests(parameters).
+
+%% Failure Tests
+
+
+%% Successful Tests
+
+test(softmax_Parameters_No_Train) :-
+ reset_Model_NoTrain,
+ parameters(PrametersList, _),
+ print('\nParameters: '),
+ print(PrametersList).
+
+test(softmax_Parameters_Direct_Input) :-
+ reset_Model_WithTrain,
+ parameters(PrametersList, _),
+ print('\nParameters: '),
+ print(PrametersList).
+
+test(softmax_Parameters_CSV_Input) :-
+ open('src/data_csv/iris2.csv', read, File),
+ take_csv_row(File, skipFirstRow,10, Data),
+ initModelWithTrain(Data, 4, [0,1,0,1,1,0,1,1,1,0], 2, 0.003, 1),
+ parameters(PrametersList, _),
+ print('\nParameters: '),
+ print(PrametersList).
+
+:- end_tests(parameters).
+
+
+%%
+%% TESTING predicate train/5
+%%
+:- begin_tests(train).
+
+%% Failure Tests
+
+test(softmax_Train_Negative_NumClass, fail) :-
+ reset_Model_NoTrain,
+ train([5.1,3.5,1.4,4.9,3.0,1.4,4.7,3.2,1.3,4.6,3.1,1.5], 3, [0,1,0,1], -1, _).
+
+test(softmax_Train_Wrong_Label_Dims1, [error(_,system_error('element-wise multiplication: incompatible matrix dimensions: 2x4 and 2x2'))]) :-
+ reset_Model_NoTrain,
+ train([5.1,3.5,1.4,4.9,3.0,1.4,4.7,3.2,1.3,4.6,3.1,1.5], 3, [0,1], 2, _).
+
+%% If the label vector is to long it seems to cause no problems
+test(softmax_Train_Wrong_Label_Dims2, [error(_,system_error('element-wise multiplication: incompatible matrix dimensions: 2x4 and 2x7'))]) :-
+ reset_Model_NoTrain,
+ train([5.1,3.5,1.4,4.9,3.0,1.4,4.7,3.2,1.3,4.6,3.1,1.5], 3, [0,1,0,1,0,0,1], 2, _).
+
+%% The same when the label values are out of range
+test(softmax_Train_Wrong_Label_Value) :-
+ reset_Model_NoTrain,
+ train([5.1,3.5,1.4,4.9,3.0,1.4,4.7,3.2,1.3,4.6,3.1,1.5], 3, [0,-1,0,-1], 2, _).
+
+%% doesnt cause a exeption
+test(softmax_Train_Too_Many_Label_Value) :-
+ reset_Model_NoTrain,
+ train([5.1,3.5,1.4,4.9,3.0,1.4,4.7,3.2,1.3,4.6,3.1,1.5], 3, [1,1,0,2], 2, _).
+
+%% doesnt cause a exeption
+test(softmax_Train_Wrong_Data_Dims) :-
+ reset_Model_NoTrain,
+ train([5.1,3.5,1.4,4.9,3.0,1.4,4.7,3.2,1.3,4.6,3.1,1.5], 4, [0,1,0], 2, _).
+
+
+%% Successful Tests
+
+test(softmax_Train_Direct_Input) :-
+ reset_Model_NoTrain,
+ train([5.1,3.5,1.4,4.9,3.0,1.4,4.7,3.2,1.3,4.6,3.1,1.5], 3, [0,1,0,1], 2, FinalValue),
+ print('\nFinalValue: '),
+ print(FinalValue).
+
+test(softmax_Train_CSV_Input) :-
+ initModelNoTrain(4, 2, 0),
+ open('src/data_csv/iris2.csv', read, File),
+ take_csv_row(File, skipFirstRow,10, Data),
+ train(Data, 4, [0,1,0,1,1,0,1,1,1,0], 2, FinalValue),
+ print('\nFinalValue: '),
+ print(FinalValue).
+
+:- end_tests(train).
-:- end_tests(predicate).
run_softmax_regression_tests :-
run_tests.
diff --git a/test_all.pl b/test_all.pl
index 570fd713d3fd70362f52b466ecf399fdaf647dd4..3ef6d61800b198bbbb237d6c811a7eee0218a158 100644
--- a/test_all.pl
+++ b/test_all.pl
@@ -6,6 +6,9 @@
%% The commented Methods either arent finished or still have some unfixed problems
+%% Make sure you run the tests in an sicstus Toplevel with prolog-mlpack-libary as its last path
+%% If not then the path to the iris2.csv will not be correct
+
:- use_module('src/methods/adaboost/adaboost_test.pl').
@@ -63,7 +66,7 @@
%%:- use_module('src/methods/random_forest/random_forest_test.pl').
-%%:- use_module('src/methods/softmax_regression/softmax_regression_test.pl').
+:- use_module('src/methods/softmax_regression/softmax_regression_test.pl').
%% better to run the sparse_coding tests alone because the c++ Method writes out alot of Debug messages that make the tests hard to read.
%%:- use_module('src/methods/sparse_coding/sparse_coding_test.pl').