Finishing demo notebook and predict_proba on unseen data

README.md

@@ -2,8 +2,6 @@ DeeSOM
 --------------------------------------------------------------------------------
 Self-organized map based classifier, developed to deal with large and highly imbalanced data.
 
-sinc(i) - http://sinc.unl.edu.ar
-
 The methods automatically build several layers of SOM. Data is clustered and samples that
 are not likely to be positive class member are discarded at each level.  
 
@@ -21,9 +19,19 @@ GPL. Please feel free to contact with any issue, comment or suggestion.
 This code was used in:
 
 "Deep neural architectures for highly imbalanced data in bioinformatics"
-L. A. Bugnon, C. Yones, D. H. Milone and G. Stegmayer*, IEEE Transactions on Neural Networks and Learning Systems,
+L. A. Bugnon, C. Yones, D. H. Milone and G. Stegmayer, IEEE Transactions on Neural Networks and Learning Systems,
  Special Issue on Recent Advances in Theory, Methodology and Applications of Imbalanced Learning (in press).
 
+sinc(i) - http://sinc.unl.edu.ar
+
+
 ## Instalation
 
-## Running the demo.
+just do:
+```bash
+python -m pip install --user -U deeSOM
+```
+
+## Running the demo
+
+You'll find a Jupyter notebook with a small tutorial to train a deeSOM model and use it for predictions.

deesom/deesom.py

@@ -287,29 +287,23 @@ def predict_proba(self, test_data=None):
         if test_data is None:
             return self.data_proba/np.max(self.data_proba)
 
-        # TODO when test_data is not none
-
-    def get_deesom_height(self, n_out, n_in):
-        """ Calculate next SOM layer height using the elastic algorithm"""
-        if self.elastic and n_out > self.elastic_threshold * n_in:
-            self.elastic_factor *= 1.2
-
-    def predict(self, test_data):
-        """
-        Implement deepsom testing per level.
-
-        """
         n_data = test_data.shape[0]
         idxtst = np.array(range(n_data))
-        slabsh = np.zeros(n_data, dtype=np.int)
+        scores = np.zeros(n_data, dtype=np.int)
         h = 0
         while h < len(self.layers) and len(idxtst) > 0:
-
             resh = self.layers[h].predict(test_data[idxtst, :])
 
             idxtst = idxtst[np.where(resh == 1)[0]]
-            slabsh[idxtst] += 1
+            scores[idxtst] += 1
             # slab[i]==n means sample i was discarded in layer n (thus it was considered as positive up to layer n-1)
             h += 1
-        return slabsh
+        return scores/np.max(scores)
+
+
+    def get_deesom_height(self, n_out, n_in):
+        """ Calculate next SOM layer height using the elastic algorithm"""
+        if self.elastic and n_out > self.elastic_threshold * n_in:
+            self.elastic_factor *= 1.2
+