Fixed IllegalStateException during models training

java/com.sap.sailing.windestimation.lab/src/com/sap/sailing/windestimation/model/classifier/maneuver/ManeuverClassifierTrainer.java

@@ -66,7 +66,6 @@ public class ManeuverClassifierTrainer {
                 LoggingUtil.logInfo("Training with  " + trainManeuvers.size() + " maneuvers...");
                 double[][] x = modelContext.getXMatrix(maneuvers);
                 int[] y = labelExtraction.getYVector(maneuvers);
-                classifierModel.resetTrainingStats();
                 classifierModel.train(x, y);
                 LoggingUtil.logInfo("Training finished. Validating on train dataset...");
                 ManeuverClassifierScoring classifierScoring = new ManeuverClassifierScoring(classifierModel);

java/com.sap.sailing.windestimation.lab/src/com/sap/sailing/windestimation/model/classifier/twdtransition/TwdTransitionClassifierTrainer.java

@@ -100,7 +100,6 @@ public class TwdTransitionClassifierTrainer {
             i++;
         }
         LoggingUtil.logInfo("Training with  " + numberOfTrainingInstances + " instances...");
-        classifierModel.resetTrainingStats();
         classifierModel.train(x, y);
         LoggingUtil.logInfo("Training finished. Validating on train dataset...");
         TwdTransitionClassifierScoring classifierScoring = new TwdTransitionClassifierScoring(classifierModel);

java/com.sap.sailing.windestimation.lab/src/com/sap/sailing/windestimation/model/regressor/twdtransition/IncrementalSingleDimensionPolynomialRegressorTrainerHelper.java

@@ -41,6 +41,9 @@ public class IncrementalSingleDimensionPolynomialRegressorTrainerHelper {
     }
 
     public void incrementRmseCalculation(double[] x, double y) {
+        if (!model.isModelReady()) {
+            model.setModelAsReadyAfterSuccessfulTraining();
+        }
         double predictedStd = model.getValue(x);
         double diff = predictedStd - y;
         squareErrorSum += diff * diff;

java/com.sap.sailing.windestimation.lab/src/com/sap/sailing/windestimation/model/regressor/twdtransition/TwdTransitionManualTrainingDataInputRegressorTrainer.java

@@ -21,8 +21,7 @@ public class TwdTransitionManualTrainingDataInputRegressorTrainer {
             throws Exception {
         IncrementalSingleDimensionPolynomialRegressorTrainerHelper trainerHelper = new IncrementalSingleDimensionPolynomialRegressorTrainerHelper(
                 regressorModelStore, model);
-        LoggingUtil
-                .logInfo("########## Training of " + model.getModelContext().getId() + " started...");
+        LoggingUtil.logInfo("########## Training of " + model.getModelContext().getId() + " started...");
         double[] modelInput = new double[1];
         for (int i = 0; i < inputOutputPairs.length; i++) {
             double xi = inputOutputPairs[i][0];

java/com.sap.sailing.windestimation/src/com/sap/sailing/windestimation/model/AbstractTrainableModel.java

@@ -51,12 +51,16 @@ public abstract class AbstractTrainableModel<InstanceType, MC extends ModelConte
 
     @Override
     public void setStatsAfterSuccessfulTraining(double trainScore, double testScore, long numberOfTrainingInstances) {
-        trainingFinished = true;
         this.trainScore = trainScore;
         this.testScore = testScore;
         this.numberOfTrainingInstances = numberOfTrainingInstances;
     }
 
+    @Override
+    public void setModelAsReadyAfterSuccessfulTraining() {
+        trainingFinished = true;
+    }
+
     @Override
     public void resetTrainingStats() {
         trainingFinished = false;

java/com.sap.sailing.windestimation/src/com/sap/sailing/windestimation/model/TrainableModel.java

@@ -46,8 +46,9 @@ public interface TrainableModel<InstanceType, MC extends ModelContext<InstanceTy
     long getNumberOfTrainingInstances();
 
     /**
-     * Sets the training statistics for this model and makes this model ready (see {@link #isModelReady()}) for
-     * prediction tasks. This method must be called only after model training has been successfully completed.
+     * Sets the training statistics for this model after successful training which should have been previously signaled
+     * with {@link #setModelAsReadyAfterSuccessfulTraining()}. This method must be called only after model training has
+     * been successfully completed.
      * 
      * @param trainScore
      *            Macro-averaged F2-Score with training data (see {@link #getTrainScore()})
@@ -58,6 +59,16 @@ public interface TrainableModel<InstanceType, MC extends ModelContext<InstanceTy
      */
     void setStatsAfterSuccessfulTraining(double trainScore, double testScore, long numberOfTrainingInstances);
 
+    /**
+     * Marks this model as ready for prediction tasks. This method should be called after successful training of this
+     * model. Afterwards, model evaluation should be performed to determine the training and test score of the model and
+     * set it with {@link #setStatsAfterSuccessfulTraining(double, double, long)}.
+     * 
+     * @see #isModelReady()
+     * @see #setStatsAfterSuccessfulTraining(double, double, long)
+     */
+    void setModelAsReadyAfterSuccessfulTraining();
+
     /**
      * Resets the training stats and marks this model as not ready (see {@link #isModelReady()}). Should be called
      * before model training starts.

java/com.sap.sailing.windestimation/src/com/sap/sailing/windestimation/model/classifier/smile/AbstractSmileClassificationModel.java

@@ -38,6 +38,7 @@ public abstract class AbstractSmileClassificationModel<InstanceType, MC extends
 
     @Override
     public void train(double[][] x, int[] y) {
+        resetTrainingStats();
         PreprocessingConfig preprocessingConfig = getPreprocessingConfig();
         scaler = null;
         if (preprocessingConfig.isScaling()) {
@@ -56,6 +57,7 @@ public abstract class AbstractSmileClassificationModel<InstanceType, MC extends
             x = pca.project(x);
         }
         internalModel = trainInternalModel(x, y);
+        setModelAsReadyAfterSuccessfulTraining();
     }
 
     protected abstract SoftClassifier<double[]> trainInternalModel(double[][] x, int[] y);