Ajout de deux trois trucs

.gitignore

@@ -1,3 +0,0 @@
-__pycache__/
-set/
-

RUN.py

@@ -3,19 +3,19 @@
 import mnist_loader
 training_data, validation_data, test_data = mnist_loader.load_data_wrapper()
 
+#print(list(training_data)[0][1])
+
 import network
-import dataset_loader
 
+net = network.Network([784, 30, 10])                                         #Testé : 94,56% / 94,87%
+net.SGD(training_data, 30, 10, 3.0, test_data=test_data)
+print("Results : {} / 10000".format(net.evaluate(test_data)))
 
-net = network.Network([262144,50, 20, 30, 10])                                         #Testé : 94,56%
-net.SGD(dataset_loader.loadTrainingSet("training"), 30, 10, 3.0, test_data=dataset_loader.loadTestSet("test"))
-
-#net = network.Network([784, 100, 10])                                          #Marche mieux apparemment
-#net.SGD(dataset_loader.loadTrainingSet("setcomplete"), 30, 10, 3.0, test_data=dataset_loader.loadTestSet("setcomplete"))
+# net = network.Network([784, 100, 10])                                          #Marche mieux apparemment
+# net.SGD(training_data, 30, 10, 3.0, test_data=test_data)
 
 # net = network.Network([784, 100, 10])                                        #Marche pas bien apparemment
 # net.SGD(training_data, 30, 10, 0.001, test_data=test_data)
 
 # net = network.Network([784, 30, 10])                                         #Marche pas du tout apparemment
 # net.SGD(training_data, 30, 10, 100.0, test_data=test_data)
-

dataset_loader.py

@@ -1,89 +0,0 @@
-from mnist_loader import load_data
-import numpy as np
-import os
-from PIL import Image
-import resource
-
-
-def vectorized_result(j):
-    """Return a 10-dimensional unit vector with a 1.0 in the jth
-    position and zeroes elsewhere.  This is used to convert a digit
-    (0...9) into a corresponding desired output from the neural
-    network."""
-    e = np.zeros((10, 1))
-    e[j] = 1.0
-    return e
-
-def loadSet(path):
-
-	filelist = []
-
-	for root, dirs, files in os.walk(path):
-		for file in files:
-			filelist.append(os.path.join(root,file))
-
-	i = 0
-	pixels = []
-	result = []
-
-
-	for name in filelist:
-
-		if i >= 500:
-
-			break
-		
-		if ".png" in name:
-
-			with Image.open(path + "/" + name.split("/")[-1]) as im:
-
-				pix = im.load()
-				temparray = []
-
-				result.append(int(name.split("/")[-1][0]))
-				
-				for x in range(im.size[0]):
-
-					for y in range(im.size[1]):
-
-						temparray.append(pix[x, y] / 255)
-
-				pixels.append(temparray)
-				print(str("%.2f" % round(i / (len(filelist) if len(filelist) < 500 else 500) * 100, 2)) + "% Done, ram usage: " + str("%.2f" % round(resource.getrusage(resource.RUSAGE_SELF).ru_maxrss / (1024*1024), 2)) + "Go", end = '\r')
-				i += 1
-
-	print("max ram usage: " + str(resource.getrusage(resource.RUSAGE_SELF).ru_maxrss / (1024*1024)) + "Go")
-
-	return (pixels, result)
-
-
-def loadTrainingSet(path):
-
-	print("importing training set...")
-
-	set = loadSet(path)
-
-	training_inputs = [np.reshape(x, (784, 1)) for x in set[0]]
-	training_results = [vectorized_result(int(y)) for y in set[1]]
-	training_data = zip(training_inputs, training_results)
-
-	return training_data
-
-def loadTestSet(path):
-
-	print("importing test set...")
-
-	set = loadSet(path)
-
-	test_inputs = [np.reshape(x, (784, 1)) for x in set[0]]
-	test_data = zip(test_inputs, set[1])
-
-	return test_data
-
-
-
-
-
-if __name__ == "__main__":
-
-	print(loadSet("set")[0])

mnist_loader.py

@@ -66,8 +66,6 @@ def load_data_wrapper():
     validation_data = zip(validation_inputs, va_d[1])
     test_inputs = [np.reshape(x, (784, 1)) for x in te_d[0]]
     test_data = zip(test_inputs, te_d[1])
-    print(te_d[0])
-    print("1:", te_d[1])
     return (training_data, validation_data, test_data)
 
 def vectorized_result(j):

network mais annotations changées.py

@@ -0,0 +1,106 @@
+import random
+import numpy as np
+
+class Network(object):
+
+    def __init__(self, sizes):
+        """sizes : [nb neurones input, nb de neurones couche 1, ..., nb de neurones couche n, nb de neurones output]
+           biases : seuils générés aléatoirement
+           weights : poids générés aléatoirement"""
+        self.num_layers = len(sizes)
+        self.sizes = sizes
+        self.biases = [np.random.randn(y, 1) for y in sizes[1:]]
+        self.weights = [np.random.randn(y, x)
+                        for x, y in zip(sizes[:-1], sizes[1:])]
+
+    def feedforward(self, a):
+        for b, w in zip(self.biases, self.weights):
+            a = sigmoid(np.dot(w, a)+b)
+        return a
+
+    def SGD(self, training_data, epochs, mini_batch_size, eta,
+            test_data=None):
+        """epochs : iterations
+           eta : taux d'apprentissage
+           test_data : s'il y en a pas le programme ne s'arrêtera pas à chaque iterations pour se tester"""
+
+        training_data = list(training_data)
+        n = len(training_data)
+
+        if test_data:
+            test_data = list(test_data)
+            n_test = len(test_data)
+
+        for j in range(epochs):
+            random.shuffle(training_data)
+            mini_batches = [
+                training_data[k:k+mini_batch_size]
+                for k in range(0, n, mini_batch_size)]
+            for mini_batch in mini_batches:
+                self.update_mini_batch(mini_batch, eta)
+            if test_data:
+                print("Epoch {} : {} / {}".format(j,self.evaluate(test_data),n_test))
+            else:
+                print("Epoch {} complete".format(j))
+
+    def update_mini_batch(self, mini_batch, eta):
+        """Met à jour les poids et seuils grâce aux gradient descendant"""
+        nabla_b = [np.zeros(b.shape) for b in self.biases]
+        nabla_w = [np.zeros(w.shape) for w in self.weights]
+        for x, y in mini_batch:
+            delta_nabla_b, delta_nabla_w = self.backprop(x, y)
+            nabla_b = [nb+dnb for nb, dnb in zip(nabla_b, delta_nabla_b)]
+            nabla_w = [nw+dnw for nw, dnw in zip(nabla_w, delta_nabla_w)]
+        self.weights = [w-(eta/len(mini_batch))*nw
+                        for w, nw in zip(self.weights, nabla_w)]
+        self.biases = [b-(eta/len(mini_batch))*nb
+                       for b, nb in zip(self.biases, nabla_b)]
+
+    def backprop(self, x, y):
+        """Calcul du gradient descendant"""
+        nabla_b = [np.zeros(b.shape) for b in self.biases]
+        nabla_w = [np.zeros(w.shape) for w in self.weights]
+        # feedforward
+        activation = x
+        activations = [x] # list to store all the activations, layer by layer
+        zs = [] # list to store all the z vectors, layer by layer
+        for b, w in zip(self.biases, self.weights):
+            z = np.dot(w, activation)+b
+            zs.append(z)
+            activation = sigmoid(z)
+            activations.append(activation)
+        # backward pass
+        delta = self.cost_derivative(activations[-1], y) * \
+            sigmoid_prime(zs[-1])
+        nabla_b[-1] = delta
+        nabla_w[-1] = np.dot(delta, activations[-2].transpose())
+        # Note that the variable l in the loop below is used a little
+        # differently to the notation in Chapter 2 of the book.  Here,
+        # l = 1 means the last layer of neurons, l = 2 is the
+        # second-last layer, and so on.  It's a renumbering of the
+        # scheme in the book, used here to take advantage of the fact
+        # that Python can use negative indices in lists.
+        for l in range(2, self.num_layers):
+            z = zs[-l]
+            sp = sigmoid_prime(z)
+            delta = np.dot(self.weights[-l+1].transpose(), delta) * sp
+            nabla_b[-l] = delta
+            nabla_w[-l] = np.dot(delta, activations[-l-1].transpose())
+        return (nabla_b, nabla_w)
+
+    def evaluate(self, test_data):
+        """Teste le programme avec le dataset fourni"""
+        test_results = [(np.argmax(self.feedforward(x)), y)
+                        for (x, y) in test_data]
+        return sum(int(x == y) for (x, y) in test_results)
+
+    def cost_derivative(self, output_activations, y):
+        """Return the vector of partial derivatives \partial C_x /
+        \partial a for the output activations."""
+        return (output_activations-y)
+
+def sigmoid(z):
+    return 1.0/(1.0+np.exp(-z))
+
+def sigmoid_prime(z):
+    return sigmoid(z)*(1-sigmoid(z))

network.py

@@ -132,8 +132,6 @@ class Network(object):
         neuron in the final layer has the highest activation."""
         test_results = [(np.argmax(self.feedforward(x)), y)
                         for (x, y) in test_data]
-
-        print(test_data[0], test_data[1])
         return sum(int(x == y) for (x, y) in test_results)
 
     def cost_derivative(self, output_activations, y):
@@ -149,6 +147,3 @@ def sigmoid(z):
 def sigmoid_prime(z):
     """Derivative of the sigmoid function."""
     return sigmoid(z)*(1-sigmoid(z))
-
-
-