KnowledgeBase is now an interface, while the implementation moved to …

…StandardKnowledgeBase. The interface contains a static factory method to produce any KB. This enable us to define the knowledgeBase field of BaseTrainable private and final.

CHANGELOG.md

@@ -1,7 +1,7 @@
 CHANGELOG
 =========
 
-Version 0.7.0-SNAPSHOT - Build 20160107
+Version 0.7.0-SNAPSHOT - Build 20160108
 ---------------------------------------
 
 - Rename the erase() method to delete() in all interfaces.
@@ -50,6 +50,7 @@ Version 0.7.0-SNAPSHOT - Build 20160107
 - Changed the behaviour and the public methods of the DatabaseConnector interface. The dropDatabase() is replaced with a clear() method that deletes all the data but does not close the connection with the database.
 - KnowledgeBase is no longer serializable. Its serializable fields are stored individually into the Database.
 - Restructuring the framework to remove all FindBug warnings.
+- KnowledgeBase is now an interface, while the implementation moved to StandardKnowledgeBase. The interface contains a static factory method to produce any KB. This enable us to define the knowledgeBase field of BaseTrainable private and final.
 
 Version 0.6.1 - Build 20160102
 ------------------------------

README.md

@@ -15,7 +15,7 @@ The code is licensed under the [Apache License, Version 2.0](https://github.com/
 Version
 -------
 
-The latest version is 0.7.0-SNAPSHOT (Build 20160107). 
+The latest version is 0.7.0-SNAPSHOT (Build 20160108). 
 
 The [master branch](https://github.com/datumbox/datumbox-framework/tree/master) is the latest stable version of the framework. The [devel branch](https://github.com/datumbox/datumbox-framework/tree/devel) is the development branch. All the previous stable versions are marked with [tags](https://github.com/datumbox/datumbox-framework/releases).
 

TODO.txt

@@ -4,6 +4,7 @@ CODE IMPROVEMENTS
 - Checkout some of the "Load of known null value" perhaps we can completely remove this parameter if it is always null. MatrixDataframe.newInstance param recordIdsReference.
 - Checkout that the example project passes as is. Check that the hybridized is on by default.
 - Update the example project to the latest version build and the config file.
+
 - Add a predictRecord() method BaseMLmodel and refactor the code to be implemented by every algorithm.
 - Add multithreading support.
 - Update all maven plugins and dependencies to their latest versions.

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

@@ -18,21 +18,23 @@
 import com.datumbox.common.dataobjects.Dataframe;
 import com.datumbox.common.objecttypes.Trainable;
 import com.datumbox.common.persistentstorage.interfaces.DatabaseConfiguration;
+import com.datumbox.framework.machinelearning.common.dataobjects.StandardKnowledgeBase;
 import com.datumbox.framework.machinelearning.common.dataobjects.KnowledgeBase;
+import java.io.Serializable;
 import java.lang.reflect.InvocationTargetException;
 import org.slf4j.Logger;
 import org.slf4j.LoggerFactory;
 
 /**
- * Base class for every Model of the Framework. This includes Machine Learning
+ * Base class for every Trainable Algorithm of the Framework. This includes Machine Learning
  * Models, Data Transformers, Feature Selectors etc.
  *
  * @author Vasilis Vryniotis <bbriniotis@datumbox.com>
  * @param <MP>
  * @param <TP>
  * @param <KB>
  */
-public abstract class BaseTrainable<MP extends BaseModelParameters, TP extends BaseTrainingParameters, KB extends KnowledgeBase<MP, TP>> implements Trainable<MP, TP> {
+public abstract class BaseTrainable<MP extends BaseModelParameters, TP extends BaseTrainingParameters, KB extends StandardKnowledgeBase<MP, TP>> implements Trainable<MP, TP> {
 
     /**
      * The Logger of all algorithms.
@@ -43,13 +45,12 @@ public abstract class BaseTrainable<MP extends BaseModelParameters, TP extends B
     /**
      * The name of the Database where we persist our data.
      */
-    protected String dbName;
+    protected final String dbName;
 
     /**
-     * The KnowledgeBase instance of the algorithm. Do NOT access it directly,
-     * use the kb() getter instead.
+     * The KnowledgeBase instance of the algorithm. 
      */
-    protected KB knowledgeBase;
+    private final KB knowledgeBase;
 
     /**
      * Generates a new instance of a BaseTrainable by providing the Class of the
@@ -58,13 +59,13 @@ public abstract class BaseTrainable<MP extends BaseModelParameters, TP extends B
      * @param <BT>
      * @param aClass
      * @param dbName
-     * @param dbConfig
+     * @param dbConf
      * @return 
      */
-    public static <BT extends BaseTrainable> BT newInstance(Class<BT> aClass, String dbName, DatabaseConfiguration dbConfig) {
+    public static <BT extends BaseTrainable> BT newInstance(Class<BT> aClass, String dbName, DatabaseConfiguration dbConf) {
         BT algorithm = null;
         try {
-            algorithm = aClass.getConstructor(String.class, DatabaseConfiguration.class).newInstance(dbName, dbConfig);
+            algorithm = aClass.getConstructor(String.class, DatabaseConfiguration.class).newInstance(dbName, dbConf);
         } 
         catch (InstantiationException | IllegalAccessException | IllegalArgumentException | InvocationTargetException | NoSuchMethodException | SecurityException ex) {
             throw new RuntimeException(ex);
@@ -74,34 +75,23 @@ public static <BT extends BaseTrainable> BT newInstance(Class<BT> aClass, String
     }
 
     /**
-     * Protected Constructor which does not include the initialization of the
-     * KnowledgeBase.
+     * The basic Constructor of all BaseTrainable classes.
      * 
-     * @param dbName
+     * @param baseDBname
      * @param dbConf 
+     * @param kbClass 
+     * @param kbSubtypeClasses 
      */
-    protected BaseTrainable(String dbName, DatabaseConfiguration dbConf) {
+    protected BaseTrainable(String baseDBname, DatabaseConfiguration dbConf, Class<? extends StandardKnowledgeBase> kbClass, Class<? extends Serializable>... kbSubtypeClasses) {
         String methodName = this.getClass().getSimpleName();
         String dbNameSeparator = dbConf.getDBnameSeparator();
-        if(!dbName.contains(methodName+dbNameSeparator)) { //patch for the K-fold cross validation which already contains the name of the algorithm in the dbname
-            dbName += dbNameSeparator + methodName;
+        if(!baseDBname.contains(methodName+dbNameSeparator)) { //patch for the K-fold cross validation which already contains the name of the algorithm in the dbname
+            baseDBname += dbNameSeparator + methodName;
         }
 
-        this.dbName = dbName;
-    }
-
-    /**
-     * Protected Constructor which includes the initialization of the
-     * KnowledgeBase.
-     * 
-     * @param dbName
-     * @param dbConf
-     * @param mpClass
-     * @param tpClass 
-     */
-    protected BaseTrainable(String dbName, DatabaseConfiguration dbConf, Class<MP> mpClass, Class<TP> tpClass) {
-        this(dbName, dbConf);
-        this.knowledgeBase = (KB) new KnowledgeBase(this.dbName, dbConf, mpClass, tpClass);
+        dbName = baseDBname;
+
+        knowledgeBase = (KB) KnowledgeBase.newInstance(kbClass, dbName, dbConf, kbSubtypeClasses);
     }
 
     /** {@inheritDoc} */

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

@@ -22,7 +22,7 @@
 
 import com.datumbox.framework.machinelearning.common.bases.baseobjects.BaseModelParameters;
 import com.datumbox.framework.machinelearning.common.bases.baseobjects.BaseTrainingParameters;
-import com.datumbox.framework.machinelearning.common.dataobjects.KnowledgeBase;
+import com.datumbox.framework.machinelearning.common.dataobjects.StandardKnowledgeBase;
 
 /**
  * Base class for all the Data Transformers of the framework.
@@ -31,7 +31,7 @@
  * @param <MP>
  * @param <TP>
  */
-public abstract class DataTransformer<MP extends DataTransformer.ModelParameters, TP extends DataTransformer.TrainingParameters> extends BaseTrainable<MP, TP, KnowledgeBase<MP, TP>> {
+public abstract class DataTransformer<MP extends DataTransformer.ModelParameters, TP extends DataTransformer.TrainingParameters> extends BaseTrainable<MP, TP, StandardKnowledgeBase<MP, TP>> {
 
     /**
      * The ModelParameters class stores the coefficients that were learned during
@@ -65,7 +65,7 @@ public static abstract class TrainingParameters extends BaseTrainingParameters {
      * @see com.datumbox.framework.machinelearning.common.bases.baseobjects.BaseTrainable#BaseTrainable(java.lang.String, com.datumbox.common.persistentstorage.interfaces.DatabaseConfiguration, java.lang.Class, java.lang.Class)  
      */
     protected DataTransformer(String dbName, DatabaseConfiguration dbConf, Class<MP> mpClass, Class<TP> tpClass) {
-        super(dbName, dbConf, mpClass, tpClass);
+        super(dbName, dbConf, StandardKnowledgeBase.class, mpClass, tpClass);
     }
 
     /**

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

@@ -22,7 +22,7 @@
 import com.datumbox.framework.machinelearning.common.bases.baseobjects.BaseTrainable;
 import com.datumbox.framework.machinelearning.common.bases.baseobjects.BaseModelParameters;
 import com.datumbox.framework.machinelearning.common.bases.baseobjects.BaseTrainingParameters;
-import com.datumbox.framework.machinelearning.common.dataobjects.KnowledgeBase;
+import com.datumbox.framework.machinelearning.common.dataobjects.StandardKnowledgeBase;
 
 /**
  * Base class for all the Feature Selectors of the framework.
@@ -31,7 +31,7 @@
  * @param <MP>
  * @param <TP>
  */
-public abstract class FeatureSelection<MP extends FeatureSelection.ModelParameters, TP extends FeatureSelection.TrainingParameters> extends BaseTrainable<MP, TP, KnowledgeBase<MP, TP>> {
+public abstract class FeatureSelection<MP extends FeatureSelection.ModelParameters, TP extends FeatureSelection.TrainingParameters> extends BaseTrainable<MP, TP, StandardKnowledgeBase<MP, TP>> {
 
     /**
      * The ModelParameters class stores the coefficients that were learned during
@@ -65,7 +65,7 @@ public static abstract class TrainingParameters extends BaseTrainingParameters {
      * @see com.datumbox.framework.machinelearning.common.bases.baseobjects.BaseTrainable#BaseTrainable(java.lang.String, com.datumbox.common.persistentstorage.interfaces.DatabaseConfiguration, java.lang.Class, java.lang.Class)  
      */
     protected FeatureSelection(String dbName, DatabaseConfiguration dbConf, Class<MP> mpClass, Class<TP> tpClass) {
-        super(dbName, dbConf, mpClass, tpClass);
+        super(dbName, dbConf, StandardKnowledgeBase.class, mpClass, tpClass);
     }
 
 

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

@@ -72,17 +72,16 @@ public static abstract class ValidationMetrics extends BaseValidationMetrics {
     }
 
     /** 
-     * @param dbName
+     * @param baseDBname
      * @param dbConf
      * @param mpClass
      * @param tpClass
      * @param vmClass
      * @param modelValidator
      * @see com.datumbox.framework.machinelearning.common.bases.baseobjects.BaseTrainable#BaseTrainable(java.lang.String, com.datumbox.common.persistentstorage.interfaces.DatabaseConfiguration, java.lang.Class, java.lang.Class)  
      */
-    protected BaseMLmodel(String dbName, DatabaseConfiguration dbConf, Class<MP> mpClass, Class<TP> tpClass, Class<VM> vmClass, ModelValidation<MP, TP, VM> modelValidator) {
-        super(dbName, dbConf);
-        this.knowledgeBase = new MLmodelKnowledgeBase<>(this.dbName, dbConf, mpClass, tpClass, vmClass);
+    protected BaseMLmodel(String baseDBname, DatabaseConfiguration dbConf, Class<MP> mpClass, Class<TP> tpClass, Class<VM> vmClass, ModelValidation<MP, TP, VM> modelValidator) {
+        super(baseDBname, dbConf, MLmodelKnowledgeBase.class, mpClass, tpClass, vmClass);
         this.modelValidator = modelValidator;
     } 
 

src/main/java/com/datumbox/framework/machinelearning/common/bases/wrappers/BaseWrapper.java

@@ -21,9 +21,9 @@
 import com.datumbox.framework.machinelearning.common.bases.baseobjects.BaseModelParameters;
 import com.datumbox.framework.machinelearning.common.bases.baseobjects.BaseTrainingParameters;
 import com.datumbox.framework.machinelearning.common.bases.mlmodels.BaseMLmodel;
-import com.datumbox.framework.machinelearning.common.dataobjects.KnowledgeBase;
 import com.datumbox.framework.machinelearning.common.bases.datatransformation.DataTransformer;
 import com.datumbox.framework.machinelearning.common.bases.featureselection.FeatureSelection;
+import com.datumbox.framework.machinelearning.common.dataobjects.StandardKnowledgeBase;
 
 /**
  * The BaseWrapper is a trainable object that uses composition instead of inheritance
@@ -35,7 +35,7 @@
  * @param <MP>
  * @param <TP>
  */
-public abstract class BaseWrapper<MP extends BaseWrapper.ModelParameters, TP extends BaseWrapper.TrainingParameters> extends BaseTrainable<MP, TP, KnowledgeBase<MP, TP>> {
+public abstract class BaseWrapper<MP extends BaseWrapper.ModelParameters, TP extends BaseWrapper.TrainingParameters> extends BaseTrainable<MP, TP, StandardKnowledgeBase<MP, TP>> {
 
     /**
      * The DataTransformer instance of the wrapper.
@@ -214,7 +214,7 @@ public void setMLmodelTrainingParameters(ML.TrainingParameters mlmodelTrainingPa
      * @see com.datumbox.framework.machinelearning.common.bases.baseobjects.BaseTrainable#BaseTrainable(java.lang.String, com.datumbox.common.persistentstorage.interfaces.DatabaseConfiguration, java.lang.Class, java.lang.Class)  
      */
     protected BaseWrapper(String dbName, DatabaseConfiguration dbConf, Class<MP> mpClass, Class<TP> tpClass) {
-        super(dbName, dbConf, mpClass, tpClass);
+        super(dbName, dbConf, StandardKnowledgeBase.class, mpClass, tpClass);
     }
 
     /** {@inheritDoc} */