Dissertation
Dissertation > Industrial Technology > Automation technology,computer technology > Automated basic theory > Artificial intelligence theory

Research and Application of Manifold-Learning Algorithms

Author ZhuHongLei
Tutor HeZuoLian
School Tianjin University
Course Applied Computer Technology
Keywords Manifold learning Search Engine Click data
CLC TP18
Type Master's thesis
Year 2007
Downloads 274
Quotes 3
Download Dissertation

In the information age , people often have to face the massive data processing , and such a large amount of data is still a geometric rate of growth . The core of these huge amounts of data , often there is a lot of redundant , so how to effectively deal with the data , find the the internal pattern between the data and effectively reduce the amount of data to extract the hidden information , the field of artificial intelligence, machine learning, data mining one of the problems . Manifold learning algorithm can effectively found that the intrinsic dimension of the high-dimensional data sets , high-dimensional data to separate the wheat from the chaff , thereby improving efficiency in the processing of vast amounts of information . This article focuses suitable for mass data fast manifold learning algorithm and its application . Mainstream manifold learning algorithm is divided into linear and nonlinear two categories . Early on the linear manifold learning algorithm PCA algorithm , easy to implement , but is only suitable for data sets with linear manifold structure ; Isomap, LLE , as the representative of non-linear manifold learning algorithm can be found nonlinear data manifolds , manifold learning algorithm time complexity is generally higher , and is not suitable for processing massive data sets . The least square error based on the anchor set isometric embedding algorithm AIE has a time complexity of O ( nlog ( n ) ) , and after the geodesic distance computation time complexity of embedded points linear , and can be fully parallel implementation AIE can effectively improve the data processing speed of the mass . Traditional search engine technology is mainly dependent on the query words entered by the user to provide search results , this method can not accurately grasp the ambiguous query words shorter case belongs to user needs , thereby reducing the quality of search results . Click data - based query expansion system , real - time capture the user clicks discrimination user needs, and AIE compressed the click data implied pages difference information , significantly reducing the space overhead of the search engine called Web difference .

Related Dissertations
More Dissertations