東華大學圖書館 |

語系: 繁體中文

說明(常見問題)

回圖書館首頁

手機版館藏查詢

登入

回首頁

切換: 標籤 | MARC模式 | ISBD

Self-organizing learning model for d...

Cai, Qiao.

FindBook

Google Book

Amazon

博客來

Self-organizing learning model for data mining applications.

紀錄類型:	書目-語言資料,印刷品 : Monograph/item
正題名/作者:	Self-organizing learning model for data mining applications./
作者:	Cai, Qiao.
面頁冊數:	179 p.
附註:	Source: Dissertation Abstracts International, Volume: 74-10(E), Section: B.
Contained By:	Dissertation Abstracts International74-10B(E).
標題:	Engineering, Computer. -
電子資源:	http://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=3568474
ISBN:	9781303230066

Self-organizing learning model for data mining applications.
Cai, Qiao.

Self-organizing learning model for data mining applications. - 179 p.

Source: Dissertation Abstracts International, Volume: 74-10(E), Section: B.

Thesis (Ph.D.)--Stevens Institute of Technology, 2013.

This dissertation is focused on improving self-organizing learning model to solve the challenging issues in data mining community. The interesting data mining problems include classification, cluster analysis, outlier detection, and dependency analysis. The basic mechanism of self-organizing learning model is derived from Kohonen's self-organizing map (SOM), which is considered as a special class of artificial neural networks based on competitive learning.

ISBN: 9781303230066Subjects--Topical Terms:

1669061
Engineering, Computer.

Self-organizing learning model for data mining applications.
LDR:03377nam a2200325 4500 001 1964244
005 20141015113819.5
008 150210s2013 ||||||||||||||||| ||eng d
020 $a 9781303230066
035 $a (MiAaPQ)AAI3568474
035 $a AAI3568474
040 $a MiAaPQ $c MiAaPQ
100 1 $a Cai, Qiao. $3 2100658
245 1 0 $a Self-organizing learning model for data mining applications.
300 $a 179 p.
500 $a Source: Dissertation Abstracts International, Volume: 74-10(E), Section: B.
500 $a Adviser: Hong Man.
502 $a Thesis (Ph.D.)--Stevens Institute of Technology, 2013.
520 $a This dissertation is focused on improving self-organizing learning model to solve the challenging issues in data mining community. The interesting data mining problems include classification, cluster analysis, outlier detection, and dependency analysis. The basic mechanism of self-organizing learning model is derived from Kohonen's self-organizing map (SOM), which is considered as a special class of artificial neural networks based on competitive learning.
520 $a The key idea of Iterative SOM with robust distance (ISOMRD) is to use the learning and clustering capabilities of SOM to obtain well-defined spatial clusters for outlier detection. Robust distance metric is used in our method to determine the adaptive threshold for identifying spatial outliers.
520 $a Incremental SOM (IncSOM) is proposed to effectively calculate the adaptive SOM grid and allow cognitive radios to manipulate the training data number and time according to the specific requirement. The advantage of IncSOM-HNN is that it can avoid the restriction of data dimension and enhance cognitive radio to better distinguish the signal types (authorized or unauthorized) in different radio scenarios.
520 $a Multiple SOMs (MSOMs) based classification methods are able to combine the advantages of both unsupervised and supervised learning mechanisms. Specifically, unsupervised SOM can search for similar properties from input data space and generate data clusters within each class, while supervised SOM can be trained from the data via label matching in the global SOM lattice space. Genetic Algorithm (GA) is used to optimize MSOMs.
520 $a Imbalanced evolving self-organizing maps (IESOM) is proposed to address the imbalanced learning problems. The SOM learning rule is modified to search the winner neuron based on energy function by minimizing local error in the competitive learning phase. The advantage of IESOM is to improve the classification performance through obtaining useful knowledge from the underrepresented minority class data.
520 $a Enhanced Dynamic Structure Preserving Map (EDSPM) is proposed to effectively improve human action recognition in video sequences. The implicit spatial temporal correlation is learned from sequential action feature sets while preserving the intrinsic topologies characterized by different human motions. A further advantage of EDSPM is its ability to project high dimensional action feature to low dimensional latent neural distribution with low computational cost.
590 $a School code: 0733.
650 4 $a Engineering, Computer. $3 1669061
690 $a 0464
710 2 $a Stevens Institute of Technology. $b Computer Engineering. $3 2049827
773 0 $t Dissertation Abstracts International $g 74-10B(E).
790 $a 0733
791 $a Ph.D.
792 $a 2013
793 $a English
856 4 0 $u http://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=3568474