half of the API changes are completed. (broken code)

TODO.txt

@@ -10,12 +10,13 @@ CODE IMPROVEMENT
 ================
 - use fit/predit/transform methods instead of training, validate etc like Python?
 	- DataTransformer: merge transform and normalize into a transform method, remove the trainingMode var and use a fit() instead
-	- FeatureSelection: change evaluateFeatures() to fit() and clearFeatures() to transform
-	- BaseMLrecommender: change train to fit()
+
+
 	- BaseMLmodel: change train(trainingData, validationData) to fit(trainingData) and a predict() method. Eliminate the test() and rename it to validate() which takes a dataset and returns its VMs
-	- Add the new methods in the BaseTrainable and the Trainable interface 
 
+- eliminate _fit() or estimateModelParameters() and pass the save() in fit()
 - Remove initializeTrainingConfiguration(), pass the parameters in train() method
+- newInstance(), constructors etc of inherited from BaseTrainable could perhaps pass to BaseTrainable and reduce the extra code.
 - Make this a method DATA_SAFE_CALL_BY_REFERENCE?
 
 - Serialization improvements

src/main/java/com/datumbox/applications/datamodeling/Modeler.java

@@ -90,10 +90,10 @@ public void train(Dataset trainingData) {
         if(selectFeatures) {
             featureSelection = FeatureSelection.newInstance(fsClass, dbName, dbConf);
             featureSelection.initializeTrainingConfiguration(trainingParameters.getFeatureSelectionTrainingParameters());
-            featureSelection.evaluateFeatures(trainingData); 
+            featureSelection.fit(trainingData); 
 
             //remove unnecessary features
-            featureSelection.clearFeatures(trainingData);  
+            featureSelection.transform(trainingData);  
         }
 
 
@@ -175,7 +175,7 @@ private BaseMLmodel.ValidationMetrics evaluateData(Dataset data, boolean estimat
             }
 
             //remove unnecessary features
-            featureSelection.clearFeatures(data);
+            featureSelection.transform(data);
         }
 
 

src/main/java/com/datumbox/applications/nlp/TextClassifier.java

@@ -141,10 +141,10 @@ public void train(Map<Object, URI> dataset) {
             featureSelection.initializeTrainingConfiguration(trainingParameters.getFeatureSelectionTrainingParameters());
 
             //find the most popular features
-            featureSelection.evaluateFeatures(trainingDataset);   
+            featureSelection.fit(trainingDataset);   
 
             //remove unnecessary features
-            featureSelection.clearFeatures(trainingDataset);
+            featureSelection.transform(trainingDataset);
         }
 
         //initialize mlmodel
@@ -274,7 +274,7 @@ public BaseMLmodel.ValidationMetrics test(Map<Object, URI> dataset) {
             }
 
             //remove unnecessary features
-            featureSelection.clearFeatures(testDataset);
+            featureSelection.transform(testDataset);
         }
 
 
@@ -347,7 +347,7 @@ private Dataset getPredictions(List<String> text) {
             }
 
             //remove unnecessary features
-            featureSelection.clearFeatures(newData);
+            featureSelection.transform(newData);
         }
 
 

src/main/java/com/datumbox/common/objecttypes/Trainable.java

@@ -56,8 +56,6 @@ public interface Trainable<MP extends Learnable, TP extends Parameterizable> {
      * @param trainingData
      * @param trainingParameters
      */
-    //public void fit(Dataset trainingData,TP trainingParameters);
-
-    public void initializeTrainingConfiguration(TP trainingParameters);
+    public void fit(Dataset trainingData, TP trainingParameters);
 
 }

src/main/java/com/datumbox/framework/machinelearning/classification/BernoulliNaiveBayes.java

@@ -78,9 +78,9 @@ public BernoulliNaiveBayes(String dbName, DatabaseConfiguration dbConf) {
     }
 
     @Override
-    public void train(Dataset trainingData, Dataset validationData) {
+    protected void _fit(Dataset trainingData) {
         knowledgeBase.getTrainingParameters().setMultiProbabilityWeighted(false);
-        super.train(trainingData, validationData);
+        super._fit(trainingData);
     }
 
     @Override

src/main/java/com/datumbox/framework/machinelearning/classification/BinarizedNaiveBayes.java

@@ -58,9 +58,9 @@ public BinarizedNaiveBayes(String dbName, DatabaseConfiguration dbConf) {
     }
 
     @Override
-    public void train(Dataset trainingData, Dataset validationData) {
+    protected void _fit(Dataset trainingData) {
         knowledgeBase.getTrainingParameters().setMultiProbabilityWeighted(false);
-        super.train(trainingData, validationData);
+        super._fit(trainingData);
     }
 
 }

src/main/java/com/datumbox/framework/machinelearning/classification/SupportVectorMachine.java

@@ -138,10 +138,11 @@ public SupportVectorMachine(String dbName, DatabaseConfiguration dbConf) {
     }
 
 
+
     @Override
-    public void train(Dataset trainingData, Dataset validationData) {
+    protected void _fit(Dataset trainingData) {
         knowledgeBase.getTrainingParameters().getSvmParameter().probability=1; //probabilities are required from the algorithm
-        super.train(trainingData, validationData);
+        super._fit(trainingData);
     }
 
     @Override

src/main/java/com/datumbox/framework/machinelearning/common/bases/BaseTrainable.java

@@ -18,6 +18,7 @@
 
 import com.datumbox.common.dataobjects.Dataset;
 import com.datumbox.common.objecttypes.Trainable;
+import com.datumbox.configuration.GeneralConfiguration;
 import com.datumbox.framework.machinelearning.common.bases.dataobjects.BaseModelParameters;
 import com.datumbox.framework.machinelearning.common.bases.dataobjects.BaseTrainingParameters;
 import com.datumbox.framework.machinelearning.common.dataobjects.KnowledgeBase;
@@ -50,24 +51,21 @@ public void erase() {
      public MP getModelParameters() {
        return knowledgeBase.getModelParameters();
 
-     }
-
-    @Override
-    public void initializeTrainingConfiguration(TP trainingParameters) {
-        //reset knowledge base
-        knowledgeBase.reinitialize();
-        knowledgeBase.setTrainingParameters(trainingParameters);
     }
 
-    /*
+
     @Override
     public void fit(Dataset trainingData, TP trainingParameters) {
+        if(GeneralConfiguration.DEBUG) {
+            System.out.println("fit()");
+        }
+
         //reset knowledge base
         knowledgeBase.reinitialize();
         knowledgeBase.setTrainingParameters(trainingParameters);
         _fit(trainingData);
     }
 
-    public abstract void _fit(Dataset trainingData);
-    */
+    protected abstract void _fit(Dataset trainingData);
+
 }

src/main/java/com/datumbox/framework/machinelearning/common/bases/datatransformation/BaseMinMaxNormalizer.java

@@ -39,15 +39,12 @@ public abstract class BaseMinMaxNormalizer extends DataTransformer<BaseMinMaxNor
     public static class ModelParameters extends DataTransformer.ModelParameters {
 
         @BigMap
-
         protected Map<Object, Double> minColumnValues;
 
         @BigMap
-
         protected Map<Object, Double> maxColumnValues;
 
         @BigMap
-
         protected Map<Object, Object> referenceLevels;
 
         public ModelParameters(DatabaseConnector dbc) {

src/main/java/com/datumbox/framework/machinelearning/common/bases/featureselection/FeatureSelection.java

@@ -56,6 +56,7 @@ public static abstract class TrainingParameters extends BaseTrainingParameters {
      * @param <F>
      * @param dbName
      * @param aClass
+     * @param dbConfig
      * @return 
      */
     public static <F extends FeatureSelection> F newInstance(Class<F> aClass, String dbName, DatabaseConfiguration dbConfig) {
@@ -84,10 +85,11 @@ protected FeatureSelection(String dbName, DatabaseConfiguration dbConf, Class<MP
     }
 
 
-    public void evaluateFeatures(Dataset trainingData) { 
+    @Override
+    protected void _fit(Dataset trainingData) { 
 
         if(GeneralConfiguration.DEBUG) {
-            System.out.println("evaluateFeatures()");
+            System.out.println("fit()");
         }
 
         estimateModelParameters(trainingData);
@@ -99,9 +101,9 @@ public void evaluateFeatures(Dataset trainingData) {
 
     }
 
-    public void clearFeatures(Dataset newData) {
+    public void transform(Dataset newData) {
         if(GeneralConfiguration.DEBUG) {
-            System.out.println("clearFeatures()");
+            System.out.println("transform()");
         }
 
         knowledgeBase.load();

src/main/java/com/datumbox/framework/machinelearning/common/bases/mlmodels/BaseMLmodel.java

@@ -137,37 +137,17 @@ public VM kFoldCrossValidation(Dataset trainingData, int k) {
      * Trains a model with the trainingData and validates it with the validationData.
      * 
      * @param trainingData
-     * @param validationData
      */
     @SuppressWarnings("unchecked")
-    public void train(Dataset trainingData, Dataset validationData) {    
-
-        if(GeneralConfiguration.DEBUG) {
-            System.out.println("train()");
-        }
-
-
+    @Override
+    protected void _fit(Dataset trainingData) {    
         if(GeneralConfiguration.DEBUG) {
             System.out.println("estimateModelParameters()");
         }
 
         //train the model to get the parameters
         estimateModelParameters(trainingData);        
 
-
-        if(validationData != null && !validationData.isEmpty()) {
-
-            if(GeneralConfiguration.DEBUG) {
-                System.out.println("validateModel()");
-            }
-
-            //validate the model with the validation data and update the validationMetrics
-            VM validationMetrics = validateModel(validationData);
-            knowledgeBase.setValidationMetrics(validationMetrics);
-
-        }
-
-
         if(GeneralConfiguration.DEBUG) {
             System.out.println("Saving model");
         }

src/main/java/com/datumbox/framework/machinelearning/common/bases/mlmodels/BaseMLrecommender.java

@@ -105,12 +105,8 @@ protected BaseMLrecommender(String dbName, DatabaseConfiguration dbConf, Class<M
      * @param trainingData
      */
     @SuppressWarnings("unchecked")
-    public void train(Dataset trainingData) {   
-
-        if(GeneralConfiguration.DEBUG) {
-            System.out.println("train()");
-        }
-
+    @Override
+    public void _fit(Dataset trainingData) {   
 
         if(GeneralConfiguration.DEBUG) {
             System.out.println("estimateModelParameters()");

src/test/java/com/datumbox/framework/machinelearning/featureselection/categorical/ChisquareSelectTest.java

@@ -95,13 +95,13 @@ public void testSelectFeatures() {
         ChisquareSelect instance = new ChisquareSelect(dbName, TestConfiguration.getDBConfig());
         instance.initializeTrainingConfiguration(param);
 
-        instance.evaluateFeatures(trainingData);
+        instance.fit(trainingData);
         instance = null;
 
 
         instance = new ChisquareSelect(dbName, TestConfiguration.getDBConfig());
 
-        instance.clearFeatures(trainingData);
+        instance.transform(trainingData);
 
         Set<Object> expResult = new HashSet<>(Arrays.asList("high_paid", "has_boat", "has_luxury_car", "has_butler", "has_pool"));
         Set<Object> result = trainingData.getColumns().keySet();

src/test/java/com/datumbox/framework/machinelearning/featureselection/categorical/MutualInformationTest.java

@@ -53,13 +53,13 @@ public void testSelectFeatures() {
         MutualInformation instance = new MutualInformation(dbName, TestConfiguration.getDBConfig());
         instance.initializeTrainingConfiguration(param);
 
-        instance.evaluateFeatures(trainingData);
+        instance.fit(trainingData);
         instance = null;
 
 
         instance = new MutualInformation(dbName, TestConfiguration.getDBConfig());
 
-        instance.clearFeatures(trainingData);
+        instance.transform(trainingData);
 
         Set<Object> expResult = new HashSet<>(Arrays.asList("high_paid", "has_boat", "has_luxury_car", "has_butler", "has_pool"));
         Set<Object> result = trainingData.getColumns().keySet();

src/test/java/com/datumbox/framework/machinelearning/featureselection/continuous/PCATest.java

@@ -57,7 +57,7 @@ public void testCalculateParameters() {
         param.setMaxDimensions(null);
         instance.initializeTrainingConfiguration(param);
 
-        instance.evaluateFeatures(originaldata);
+        instance.fit(originaldata);
         instance=null;
 
         Dataset newdata = originaldata;
@@ -72,7 +72,7 @@ public void testCalculateParameters() {
         expResult.add(Record.<Double>newDataVector(new Double[]{-14.1401, -6.4677, 1.4920}, null));
         expResult.add(Record.<Double>newDataVector(new Double[]{-23.8837, 3.7408, -2.3614}, null));
 
-        instance.clearFeatures(newdata);
+        instance.transform(newdata);
 
         assertEquals(newdata.size(), expResult.size());
 

src/test/java/com/datumbox/framework/machinelearning/featureselection/scorebased/TFIDFTest.java

@@ -80,13 +80,13 @@ public void testSelectFeatures() {
         TFIDF instance = new TFIDF(dbName, TestConfiguration.getDBConfig());
         instance.initializeTrainingConfiguration(param);
 
-        instance.evaluateFeatures(trainingData);
+        instance.fit(trainingData);
         instance = null;
 
 
         instance = new TFIDF(dbName, TestConfiguration.getDBConfig());
 
-        instance.clearFeatures(trainingData);
+        instance.transform(trainingData);
 
         Set<Object> expResult = new HashSet<>(Arrays.asList("important1", "important2", "important3"));
         Set<Object> result = trainingData.getColumns().keySet();

src/test/java/com/datumbox/framework/machinelearning/recommendersystem/CollaborativeFilteringTest.java

@@ -159,7 +159,7 @@ public void testPredict() {
         CollaborativeFiltering.TrainingParameters param = new CollaborativeFiltering.TrainingParameters();
         param.setSimilarityMethod(CollaborativeFiltering.TrainingParameters.SimilarityMeasure.PEARSONS_CORRELATION);
         instance.initializeTrainingConfiguration(param);
-        instance.train(trainingData);
+        instance._fit(trainingData);
 
 
         instance = null;

src/test/java/com/datumbox/framework/machinelearning/regression/NLMSTest.java

@@ -182,14 +182,14 @@ public void testKFoldCrossValidation() {
         featureSelectionParameters.setWhitened(false);
         featureSelectionParameters.setVarianceThreshold(0.99999995);
         featureSelection.initializeTrainingConfiguration(featureSelectionParameters);
-        featureSelection.evaluateFeatures(trainingData);
+        featureSelection.fit(trainingData);
         /*
         featureSelection=null;
         
         
         featureSelection = new PCA(dbName, TestConfiguration.getDBConfig());
         */
-        featureSelection.clearFeatures(trainingData);
+        featureSelection.transform(trainingData);
         featureSelection.erase();