Draft HepNet class

nnfwtbn/model.py

@@ -32,8 +32,6 @@ class CrossValidator(ABC):
         # Handle variable
         if isinstance(self.variable, str):
             self.variable = Variable(self.variable, self.variable)
-                
-
 
     @abstractmethod
     def select_slice(self, df, slice_id):
@@ -202,33 +200,28 @@ class HepNet:
         Creates a new HEP model. The keras model parameter is a callable that
         returns a new instance of the model (The HEP net needs to able to create
         multiple models, one for each cross validation fold.)
-        """
 
-    def set(self, key, value):
-        """
-        Stores properties such as normalization moments.
-        """
+        The cross_validator must be a CrossValidator object.
 
-    def get(self, key, value):
-        """
-        Returns properties such as normalization moments.
+        The normalizor must be a callable that returns a normalizor. Each
+        cross_validation fold uses a separate normalizor with independent
+        normalization weights.
         """
+        self.model = keras_model
+        self.cv = cross_validator
+        self.norm = normalizor
 
-    def store(self, file):
-        """
-        Write properties and training weights to the given file.
-        """
-
-    def restore(self, file):
-        """
-        Loads properties and training weights from the given file.
-        """
-    
     def fit(self, df, selection):
         """
-        Calls fit() on the Keras model
+        Calls fit() on all folds.
         """
 
+        ### Loop over folds:
+
+        # seed normalizors
+
+        # fit folds
+
     def predict(self, df):
         """
         Calls predict() on the Keras model