| In recent years, a lot of people working on the field of Machine Learning, especially on Deep Learning have a huge progress. In this field people argue how to build multilayer’s artificial neutral architecture makes the deep learning have a better way be used on the machine learning problems. On the other side, the architecture of deep learning has a similar action on handling information compared with other methods. Because of that, someone confirms that deep learning may be is a breakthrough to solve the complex Artificial Intelligence.At present in deep learning the most efficiently way to do some research is Deep Brief Network(DBN). Because of the DBN is easy to learning compare with other method, and it overcome efficient of train the multilayer directly. It can be seen that DBN is a very valuable method to have a research in the field of deep learning.At the same time, the Topic Model is a probability model which appeared in the text mining, it isn’t like the traditional model just thinking about the text on the dimension of dictionary. It add a dimension on to it, and makes the text can have a probability model on the topic. That means we will have many benefits. First, it reduce the dimensions, that means the curse of dimensionality will not happen, because of that we can see the topic model is also a very important model.In our research we will continue to argue the problem of the DBN. Our main contribution is:We summarized the progress learning algorithm of DBN and discovered the DBN can used with topic model which creat a new model to have a text classification.With the development of deep learning, I’m sure that our method will become more and more important, and more people will do more research on this method. And with the research on text classification and deep learning get more and more, we believe that our method will be more valuable. Our research first propose this method, breakthrough a possible way of this new direction. |
PDF Full Text Request