init

connection.py

@@ -0,0 +1,23 @@
+import random
+
+class Connection:
+    def __init__(self, in_, out_):
+        self.weight = random.uniform(-1.0, 1.0) 
+        self.in_ = in_
+        self.out_ = out_
+        self.value = 0.0
+
+    def transfer(self, value:float):
+        self.value = value
+
+    def adjust(self, delta:float, learning_rate:float):
+        old_weight = self.weight
+        
+        gradient = delta * self.value
+        self.weight += learning_rate * gradient
+
+        if self.in_ is None:
+            return
+
+        error_for_prev = delta * old_weight
+        self.in_.adjust(error_for_prev, learning_rate)

data.train

@@ -0,0 +1,38 @@
+[
+    [[1,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0], [0, 1, 0]],
+    [[0,1,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0], [1, 0, 0]],
+    [[0,0,1,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0], [1, 0, 0]],
+    [[0,0,0,1,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0], [1, 0, 0]],
+    [[0,0,0,0,1,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0], [0, 1, 0]],
+    [[0,0,0,0,0,1,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0], [1, 0, 0]],
+    [[0,0,0,0,0,0,1,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0], [1, 0, 0]],
+    [[0,0,0,0,0,0,0,1,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0], [1, 0, 0]],
+    [[0,0,0,0,0,0,0,0,1,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0], [0, 1, 0]],
+    [[0,0,0,0,0,0,0,0,0,1,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0], [1, 0, 0]],
+    [[0,0,0,0,0,0,0,0,0,0,1,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0], [1, 0, 0]],
+    [[0,0,0,0,0,0,0,0,0,0,0,1,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0], [1, 0, 0]],
+    [[0,0,0,0,0,0,0,0,0,0,0,0,1,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0], [1, 0, 0]],
+    [[0,0,0,0,0,0,0,0,0,0,0,0,0,1,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0], [1, 0, 0]],
+    [[0,0,0,0,0,0,0,0,0,0,0,0,0,0,1,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0], [0, 1, 0]],
+    [[0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,1,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0], [1, 0, 0]],
+    [[0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,1,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0], [1, 0, 0]],
+    [[0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,1,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0], [1, 0, 0]],
+    [[0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,1,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0], [1, 0 ,0]],
+    [[0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,1,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0], [1, 0, 0]],
+    [[0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,1,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0], [0, 1, 0]],
+    [[0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,1,0,0,0,0,0,0,0,0,0,0,0,0,0,0], [1, 0, 0]],
+    [[0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,1,0,0,0,0,0,0,0,0,0,0,0,0,0], [1, 0, 0]],
+    [[0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,1,0,0,0,0,0,0,0,0,0,0,0,0], [1, 0, 0]],
+    [[0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,1,0,0,0,0,0,0,0,0,0,0,0], [1, 0, 0]],
+    [[0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,1,0,0,0,0,0,0,0,0,0,0], [1, 0, 0]],
+    [[0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,1,0,0,0,0,0,0,0,0,0], [0, 0, 1]],
+    [[0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,1,0,0,0,0,0,0,0,0], [0, 0, 1]],
+    [[0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,1,0,0,0,0,0,0,0], [0, 0, 1]],
+    [[0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,1,0,0,0,0,0,0], [0, 0, 1]],
+    [[0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,1,0,0,0,0,0], [0, 0, 1]],
+    [[0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,1,0,0,0,0], [0, 0, 1]],
+    [[0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,1,0,0,0], [0, 0, 1]],
+    [[0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,1,0,0], [0, 0, 1]],
+    [[0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,1,0], [0, 0, 1]],
+    [[0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,1], [0, 0, 1]]
+]

main.py

@@ -0,0 +1,28 @@
+import string
+from network import Network
+from trainer import Trainer
+
+network = Network()
+network.build_layers(3, 36, 3, 12)
+network.connect_layers()
+
+trainer = Trainer(network, "./data.train")
+trainer.train(20000)
+
+print("Erkenne die Art eines zeichens (Konsontant, Vokal oder Zahl).")
+while True:
+    inpt = input("Dein Zeichen: ").lower()
+
+    vec = [0 for _ in range(0, 36)]
+    refrence = string.ascii_lowercase + string.digits
+    vec[refrence.index(inpt)] = 1
+
+    output = network.fire(vec)
+
+    print(f"Konsontant: {round(output[0], 2)} | Vokal: {round(output[1], 2)} | Zahl: {round(output[2], 2)}")
+    if max(output) == output[0]:
+        print("AI Denkt: Konsontant")
+    elif max(output) == output[1]:
+        print("AI Denkt: Vokal")
+    else:
+        print("AI Denkt: Zahl")

network.py

@@ -0,0 +1,49 @@
+from neuron import Neuron
+from connection import Connection
+import random
+import json
+
+class Network():
+    def __init__(self):
+        self.layers = []
+
+    def build_layers(self, count:int, input_size:int, output_size:int, layer_size:int):
+
+        for layer in range(0, count):
+            self.layers.append([])
+
+            for _ in range(0, input_size if layer == 0 else output_size if layer == count - 1 else layer_size):
+                self.layers[layer].append(Neuron())
+
+
+    def _connect_layer(self, last_layer, current_layer):
+        for last_neuron in last_layer:
+            for current_neuron in current_layer:
+                connection = Connection(last_neuron, current_neuron)
+                last_neuron.connect_out(connection)
+                current_neuron.connect_in(connection)
+
+    def connect_layers(self):
+
+        for i in range(1, len(self.layers)):
+            last_layer = self.layers[i-1]
+            current_layer = self.layers[i]
+
+            self._connect_layer(last_layer, current_layer)
+
+    def fire(self, vec):
+
+        for l in range(0, len(self.layers)):
+            layer = self.layers[l]
+            for i in range(0, len(layer)):
+                neuron = layer[i]
+                neuron.calc(vec[i] if l == 0 else None)
+
+        output = []
+
+        for i in range(0, len(self.layers[-1])):
+            layer = self.layers[-1]
+            output.append(layer[i].value)
+
+        return output
+

neuron.py

@@ -0,0 +1,39 @@
+import random
+import math
+
+class Neuron:
+    def __init__(self):
+        self.bias = random.uniform(-1.0, 1.0)
+        self.in_ = []
+        self.out_ = []
+        self.value = 0.0
+
+    def connect_in(self, connection):
+        self.in_.append(connection)
+
+    def connect_out(self, connection):
+        self.out_.append(connection)
+
+    def calc(self, input_val=None):
+        if input_val is not None:
+            self.value = input_val
+        else:
+            z = self.bias
+            for connection in self.in_:
+                z += connection.value * connection.weight
+            
+            z = max(-500, min(500, z)) 
+            self.value = 1.0 / (1.0 + math.exp(-z))
+
+        for connection in self.out_:
+            connection.transfer(self.value)
+
+    def adjust(self, error:float, learning_rate:float = 0.1):
+        derivative = self.value * (1.0 - self.value)
+        
+        delta = error * derivative
+
+        self.bias += learning_rate * delta
+
+        for connection in self.in_:
+            connection.adjust(delta, learning_rate)

trainer.py

@@ -0,0 +1,32 @@
+from network import Network
+import json
+import random
+
+class Trainer:
+    def __init__(self, network: Network, train_file):
+        self.network = network
+
+        with open(train_file, "r") as file:
+            f_data = file.read()
+            self.data = json.loads(f_data)
+
+    def adjust(self, check_vec, layer_idx):
+        layer = self.network.layers[layer_idx]
+
+        for i, neuron in enumerate(layer):
+            error = check_vec[i] - neuron.value
+            neuron.adjust(error)
+
+    def train(self, iters):
+        for i in range(0, iters):
+            print(f"Train {(i / iters) * 100}%")
+            sample = random.choice(self.data)
+            in_vec = sample[0]
+            check_vec = sample[1]
+
+            self.network.fire(in_vec)
+
+            self.adjust(check_vec, len(self.network.layers) - 1)
+
+
+