Introduction to Learning sorting algorithms

Introduction to Learning sorting algorithms

Learning sequencing (learning to Rank, LTR) is a sort of algorithm based on machine learning method.

Traditional classical models, such as the VSM model based on the TFIDF feature, are difficult to incorporate into many features, that is, in addition to the TFIDF features, they cannot be incorporated into other types of features.

The method of machine learning is easy to fuse many features, and has a mature and deep theoretical basis, the parameters are calculated by iteration, a set of mature theory to solve the problem of sparse, overfitting and so on.

The Ltr method can be broadly divided into three categories:

1) Pointwise Single Document method

2) pairwise Document pair method

3) Listwise Document List method

The LTR algorithm has many applications in practice, such as you enter a query in the search engine to get a series of search results, so how do you sort the results according to your query? Because we need to be the most relevant to query the result of the first bar ~ Machine translation will also have a translation of the results of the sequencing process. So, the LTR algorithm is still very useful.
In the following blog, we will specifically describe each of the above algorithms ~

http://blog.csdn.net/nanjunxiao/article/details/8976195

Introduction to Learning sorting algorithms