Define adapters for machine learning models used in reconstruction.

cml/ports/ml_adapter.py

 from collections import Counter
-from abc import ABCMeta, abstractmethod
+from abc import abstractmethod
 
 from numpy import array, linspace, less, greater, std, argsort
 from scipy.signal import argrelextrema
 import sklearn.cluster
 from sklearn.neighbors.kde import KernelDensity
+from sklearn.metrics import max_error, mean_absolute_error
 from keras.layers import Input, Dense
 from keras.models import Model
 from keras.regularizers import l1
@@ -94,6 +95,49 @@ class ConstructionClusteringMLModel(MachineLearningModel):
         return self
 
 
+class ReconstructionConceptualMLModel(MachineLearningModel):
+
+    def __init__(self, model):
+        self.__model = model
+        self.accuracy = None
+
+    def train(self, data, *args, **kwargs):
+        # TODO (dmt): Improve signature of this function!
+        labels = args[0]
+        self.__model = self.__model.fit(data, labels)
+        self.accuracy = self.__model.score(data, labels)
+        return self
+
+    def predict(self, data):
+        return [i for i in self.__model.predict(data)]
+
+
+class ReconstructionProceduralMLModel(MachineLearningModel):
+    def __init__(self, model):
+        self.__model = model
+        self.mean_error = None
+        self.max_error = None
+
+    def train(self, data, *args, **kwargs):
+        # TODO (dmt): Provide a better way dealing with
+        # zero values as max_abs_label!
+        labels = args[0]
+        self.__model = self.__model.fit(data, labels)
+        relative_max_error = max_error(y_true=labels,
+                                       y_pred=self.__model.predict(data))
+        max_abs_label = max((abs(i) for i in labels))
+        if max_abs_label == 0:
+            raise ValueError()
+        self.max_error = (relative_max_error*100)/max_abs_label
+        relative_mean_error = mean_absolute_error(
+            y_true=labels, y_pred=self.__model.predict(data))
+        self.mean_error = (relative_mean_error*100)/max_abs_label
+        return self
+
+    def predict(self, data):
+        return [i for i in self.__model.predict(data)]
+
+
 class KernelDensityEstimator(MachineLearningModel):
 
     def __init__(self, kernel="gaussian", bandwidth=3, gridsize=256):
@@ -107,8 +151,8 @@ class KernelDensityEstimator(MachineLearningModel):
         if not self.__model:
             self.__model = KernelDensity(kernel=self.kernel,
                                          bandwidth=self.bandwidth)
-
         self.__model.fit(reshaped_data)
+
         return self
 
     def density(self):