Multiple Changes to the CNN implementation

Ki/Core/ConvolutionNetwork.py

@@ -17,56 +17,58 @@
 
 import numpy as np
 
-def ConvolutionNetwork():
+class ConvolutionNetwork():
 
     Layers = []
     LayerTypes = []
     Activations = []
     NumberOutputs = 0
+    InputSize = []
     Fitness = 0
 
     #get the hyper parameters ready
-    def __init__(self, _LayerSizes, _LayerTypes, _NumberOutputs):
+    def __init__(self, _LayerSizes, _LayerTypes, _InputSize, _NumberOutputs):
 
         self.Layers = []
         self.LayerTypes = []
         self.Activations = []
         self.Fitness = 0
         self.NumberOutputs = _NumberOutputs
+        self.InputSize = _InputSize
 
         self.CreateNetwork(_LayerSizes, _LayerTypes)
 
-    #need to make this more dynamic...
-    def CreateNetwork(self, _LayerSizes, _LayerTypes):
-
-        Layer1 = []
-        Layer2 = []
-        Layer3 = []
-        Layer4 = []
-        Layer5 = []
-        Layer6 = []
+    def AddLayers(self, FilterSize=[3,3], NumberFilters=8, NumberLayers=2):
 
-        for i in range(32):
+        for i in range(NumberLayers):
+            Layer = []
 
-            Layer1.append(Filter())
-
-        Layer2.append(Pooling())
+            for i in range(NumberFilters):
+                Layer.append(Filter(FilterSize))
 
-        for i in range(32):
-
-            Layer3.append(Filter())
+        self.Layers.append(Layer)
+        self.LayerTypes.append(0)
 
-        Layer4.append(Pooling())
+        Layer = [Pooling()]
+        self.Layers.append(Layer)
+        self.LayerTypes.append(1)
+
+    #Dynamically add Filter and Pooling Layers
+    def CreateNetwork(self, _LayerSizes, _LayerTypes):
 
-        for i in range(self.NumberOutputs):
-
-            Layer6.append(Node())
+        for i in range(2):
+            self.AddLayers([3+i,3+i], 16, 1)
 
 
 
+    def CalculateNumberNeurons(self):
+
+        zeros = np.zeros(self.InputSize)
+
+        #for i in range(self.LayerTypes):
+
 
-
-
+CNN = ConvolutionNetwork(0,0, [256,256], 3) 
 
 
 

Ki/Core/Filter.py

@@ -17,13 +17,14 @@ class Filter():
     FilterSize = []
     Stride = []
     Padding = []
+    OutData = []
 
     def __init__(self, _FilterSize=[3,3], _Stride=[1,1], _Padding=[1,1,1,1]):
 
         self.FilterSize = _FilterSize
         self.Stride = _Stride
         self.Padding = _Padding
-
+        self.OutData = []
         self.Filter = np.random.uniform(low =-1, high=1, size=[self.FilterSize[0], self.FilterSize[1]])
 
     #Apply a filter to some input data
@@ -53,7 +54,7 @@ def ApplyActivation(self, Data):
         for i in range(Data.shape[0]):
             for j in range(Data.shape[1]):
                 Data[i,j] = a.Relu(Data[i,j])
-
+'''
 #Test Data of a Cube
 data = [[1,1,1,1,1,1,1,1,1],
         [1,0,0,0,0,0,0,0,1],
@@ -84,7 +85,6 @@ def ApplyActivation(self, Data):
 
 
 
-
-
+'''
 
 

Ki/Core/Pooling.py

@@ -9,10 +9,12 @@
 class Pooling():
 
     WindowSize = [2,2]
+    OutData = []
 
     def __init__(self, _WindowSize=[2,2]):
 
         self.WindowSize = _WindowSize
+        self.OutData = []
 
     def MaxPoolData(self, Data):
 
@@ -35,7 +37,7 @@ def MaxPoolData(self, Data):
     def Flatten(self, data):
 
         return np.ravel(data)
-        
+'''
 #Test Data of a Cube
 data = [[1,1,1,1,1,1,1,1,1],
         [1,0,0,0,0,0,0,0,1],
@@ -54,5 +56,5 @@ def Flatten(self, data):
 pooled2 = pool.MaxPoolData(pooled)
 flat = pool.Flatten(pooled2)
 
-
+'''