自编码算法与稀疏性

From Ufldl

Jump to: navigation, search
Line 4: Line 4:
二审: 新浪微博,@大黄蜂的思索 http://weibo.com/u/1733291480
二审: 新浪微博,@大黄蜂的思索 http://weibo.com/u/1733291480
 +
 +
【原文】
 +
 +
So far, we have described the application of neural networks to supervised learning, in which we have labeled
 +
training examples.  Now suppose we have only a set of unlabeled training examples <math>\textstyle \{x^{(1)}, x^{(2)}, x^{(3)}, \ldots\}</math>,
 +
where <math>\textstyle x^{(i)} \in \Re^{n}</math>.  An
 +
'''autoencoder''' neural network is an unsupervised learning algorithm that applies backpropagation,
 +
setting the target values to be equal to the inputs.  I.e., it uses <math>\textstyle y^{(i)} = x^{(i)}</math>.
 +
 +
Here is an autoencoder:
 +
 +
【初译】
 +
 +
目前为止,我们已经讨论了神经网络在监督学习中的应用。在监督学习中,训练样本是有类别标签的。现在假设我们只有一个没有类别标签的训练样本集合 <math>\textstyle \{x^{(1)}, x^{(2)}, x^{(3)}, \ldots\}</math> ,其中 <math>\textstyle x^{(i)} \in \Re^{n}</math> 。一个自编码神经网络是一种非监督学习算法,它使用了反向传播算法,并将目标值设为输入值,比如 <math>\textstyle y^{(i)} = x^{(i)}</math> 。下图是一个自编码神经网络的示例。

Revision as of 11:30, 12 March 2013

Personal tools