update dA

Author: Syndrome777

5_denoising_autoencoders_降噪自动编码.md

@@ -219,9 +219,31 @@ class dA(object):
 
         return (cost, updates)
 ```
+我们现在可以定义一个函数来实现重复的更新参数W，b，b‘，直到这个重构消耗大约是最小的。
 
-
-
+```Python
+    da = dA(
+        numpy_rng=rng,
+        theano_rng=theano_rng,
+        input=x,
+        n_visible=28 * 28,
+        n_hidden=500
+    )
+
+    cost, updates = da.get_cost_updates(
+        corruption_level=0.,
+        learning_rate=learning_rate
+    )
+
+    train_da = theano.function(
+        [index],
+        cost,
+        updates=updates,
+        givens={
+            x: train_set_x[index * batch_size: (index + 1) * batch_size]
+        }
+    )
+```
 
 
 

README.md

@@ -17,7 +17,7 @@ This is a `Chinese tutorial` which is translated from [DeepLearning 0.1 document
 * [使用逻辑回归进行MNIST分类(Classifying MNIST digits using Logistic Regression)](https://github.com/Syndrome777/DeepLearningTutorial/blob/master/2_Classifying_MNIST_using_LR_逻辑回归进行MNIST分类.md)
 * [多层感知机（Multilayer Perceptron）](https://github.com/Syndrome777/DeepLearningTutorial/blob/master/3_Multilayer_Perceptron_多层感知机.md)
 * [卷积神经网络（Convolutional Neural Networks(LeNet)）](https://github.com/Syndrome777/DeepLearningTutorial/blob/master/4_Convoltional_Neural_Networks_LeNet_卷积神经网络.md)
-* Denoising Autoencoders(dA)
+* [降噪自动编码机器（Denoising Autoencoders(dA)）](https://github.com/Syndrome777/DeepLearningTutorial/blob/master/5_Denoising_Autoencoders_降噪自动编码.md)
 * Stcaked Denoising Autoencoders(SdA)
 * Restricted Boltzmann Machines(RBM)
 * Deep Belif Networks