質問編集履歴

コードの全文を付け足しました。

2018/10/22 14:52

投稿

knght

スコア14

test CHANGED Viewed

File without changes

test CHANGED Viewed

@@ -73,3 +73,269 @@
 print('accuracy:', accuracy)
 ```
+クラス内のコードの全文です。
+```Tensorflow
+# activation = [0, 1, 0 ,1]
+# 0 : ReLU, 1 : sigmoid
+class MLP(object):
+    def __init__(self, input_n, output_n, hidden_n, func_n):
+        self.input_n = input_n
+        self.output_n = output_n
+        self.hidden_n = hidden_n
+        self.weights = []
+        self.biases = []
+        self.func_n = func_n
+        self._x = None
+        self._t = None
+        self._sess = None
+        self._history = {
+            'accuracy' : [],
+            'loss' : []
+        }
+    def weight_variable(self, shape):
+        initial = tf.Variable(tf.random_normal(shape, stddev=0.01))
+        return initial
+    def bias_variable(self, shape):
+        initial = tf.Variable(tf.zeros(shape))
+        return initial
+    def function(self, x, func):
+        if func == 0:
+            return tf.nn.relu(x)
+        if func == 1:
+            return tf.nn.sigmoi(x)
+    def inference(self, x):
+        # input-layer - hidden-layer, hidden-layer - hidden-layer
+        for i, hidden in enumerate(self.hidden_n):
+            if i == 0:
+                inp = x
+                input_dim = self.input_n
+            else:
+                inp = output
+                input_dim = self.hidden_n[i-1]
+            self.weights.append(self.weight_variable([input_dim, hidden]))
+            self.biases.append(self.bias_variable([hidden]))
+            h = tf.matmul(inp, self.weights[-1]) + self.biases[-1]
+            output = self.function(h, self.func_n[i])
+        self.weights.append(self.weight_variable([self.hidden_n[-1], self.output_n]))
+        self.biases.append(self.bias_variable([self.output_n]))
+        y = tf.nn.softmax(tf.matmul(output, self.weights[-1]) + self.biases[-1])
+        return y
+    def loss(self, y, t):
+        loss = tf.reduce_mean(-tf.reduce_sum(
+            t * tf.log(tf.clip_by_value(y, 1e-10, 1.0)), reduction_indices=[1]))
+        return loss
+    def training(self, loss):
+        optimizer = tf.train.GradientDescentOptimizer(0.01)
+        train_step = optimizer.minimize(loss)
+        return(loss)
+    def accuracy(self, y, t):
+        correct_prediction = tf.equal(tf.argmax(y, 1), tf.argmax(t, 1))
+        accuracy = tf.reduce_mean(tf.cast(correct_prediction, tf.float32))
+        return accuracy
+    def fit(self, X_train, Y_train, epochs=100, batch_size=100):
+        x = tf.placeholder(tf.float32, [None, self.input_n])
+        t = tf.placeholder(tf.float32, [None, self.output_n])
+        # save for evaluate()
+        self._x = x
+        self._t = t
+        y = self.inference(x)
+        loss = self.loss(y, t)
+        train_step = self.training(loss)
+        accuracy = self.accuracy(y, t)
+        init = tf.global_variables_initializer()
+        sess = tf.Session()
+        sess.run(init)
+        # save for evaluate()
+        self._sess = sess
+        N_train = len(X_train)
+        n_batches = N_train // batch_size
+        for epoch in range(epochs):
+            X_, Y_ = shuffle(X_train, Y_train)
+            for i in range(n_batches):
+                start = i * batch_size
+                end = start + batch_size
+                sess.run(train_step, feed_dict={
+                    x : X_train[start:end],
+                    t : Y_train[start:end]
+                })
+            loss_ = loss.eval(session=sess, feed_dict={
+                x : X_train,
+                t : Y_train
+            })
+            accuracy_ = accuracy.eval(session=sess, feed_dict={
+                x : X_train,
+                t : Y_train
+            })
+            # save to _history
+            self._history['loss'].append(loss_)
+            self._history['accuracy'].append(accuracy_)
+        return self._history
+    def evaluate(self, X_test, Y_test):
+        return self.accuracy.eval(session=self._sess, feed_dict={
+            self._x : X_test,
+            self._t : Y_test
+        })
+```