Added ai that works i think

rgbAI/lib/func.py

@@ -6,7 +6,7 @@ class AIlib:
         return 1/(1 + np.exp(-x))
 
     def correctFunc(inp:np.array): # generates the correct answer for the AI 
-        return np.array( [inp[2], inp[1], inp[0]] ) # basically invert the rgb values
+        return np.asarray( [1.0 - inp[0], 1.0 - inp[1], 1.0 - inp[2]] ) # basically invert the rgb values
 
     def calcCost( predicted:np.array, correct:np.array ): # cost function, lower -> good, higher -> bad, bad bot, bad
         costSum = 0
@@ -87,15 +87,27 @@ class AIlib:
             obj.weights[i] -= obj.learningrate * gradient[i]["weight"] # mutate the weights
             obj.bias[i] -= obj.learningrate * gradient[i]["bias"]
 
-    def learn( inp:np.array, obj, theta:float ):
+    def learn( inputNum:int, targetCost:float, obj, curCost: float=None ):
         # Calculate the derivative for:
         # Cost in respect to weights
         # Cost in respect to biases
 
         # i.e. : W' = W - lr * gradient (respect to W in layer i) = W - lr*[ dC / dW[i] ... ]
         # So if we change all the weights with i.e. 0.01 = theta, then we can derive the gradient with math and stuff
-        maxLen = len(obj.bias)
-        grads, res, cost = AIlib.gradient( inp, obj, theta, maxLen - 1 )
-        AIlib.mutateProps( obj, maxLen, grads ) # mutate the props for next round
 
-        print("Cost:", cost, "|", inp, res)
+        if( not curCost or curCost > targetCost ): # targetCost is the target for the cost function
+            inp = np.asarray(np.random.rand( 1, inputNum ))[0]
+
+            maxLen = len(obj.bias)
+            grads, res, curCost = AIlib.gradient( inp, obj, theta, maxLen - 1 )
+            AIlib.mutateProps( obj, maxLen, grads ) # mutate the props for next round
+            print("Cost:", curCost, "|", inp, res)
+
+            return AIlib.learn( inputNum, targetCost, obj, theta, curCost )
+
+        else:
+            print("DONE\n")
+            print(obj.weights)
+            print(obj.bias)
+            return
+

rgbAI/main.py

@@ -14,9 +14,7 @@ class rgb(object):
             self.bias = [ ai.genRandomMatrix(1, 4), ai.genRandomMatrix(1, 4), ai.genRandomMatrix(1, 3) ]
             # This doesn't look very good, but it works so...
 
-            self.generation = 0
-
-            self.learningrate = 0.1 # the learning rate of this ai
+            self.learningrate = 0.01 # the learning rate of this ai
 
             print( self.weights )
             print( self.bias )
@@ -30,8 +28,8 @@ class rgb(object):
         # Cost needs to get to 0, we can figure out this with backpropagation
         return cost
 
-    def learn( self, inp:np.array, theta:float ):
-        ai.learn( inp, self, theta )
+    def learn( self ):
+        ai.learn( 3, 0.001, self, 0.0001 )
 
     def think( self, inp:np.array ):
         print("-----Gen " + str(self.generation) + "------")
@@ -46,17 +44,8 @@ class rgb(object):
         print("\n----------------\n\n")
         return res
 
-    def train( self ):
-        for i in range(self.traintimes):
-            inpArr = np.asarray(np.random.rand( 1, 3 ))[0]
-            self.generation = i
-            self.learn( inpArr, 0.1 )
-
 def init():
     bot = rgb()
-
-    bot.traintimes = 10000
-    bot.train()
-
+    bot.learn()
 
 init()