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

An algorithm for fitting the contras...

Bearden, J. Neil.

FindBook

Google Book

Amazon

博客來

An algorithm for fitting the contrast model.

紀錄類型:	書目-電子資源 : Monograph/item
正題名/作者:	An algorithm for fitting the contrast model./
作者:	Bearden, J. Neil.
面頁冊數:	68 p.
附註:	Source: Dissertation Abstracts International, Volume: 65-07, Section: B, page: 3767.
Contained By:	Dissertation Abstracts International65-07B.
標題:	Psychology, Psychometrics. -
電子資源:	http://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=3140284
ISBN:	049687425X

An algorithm for fitting the contrast model.
Bearden, J. Neil.

An algorithm for fitting the contrast model. - 68 p.

Source: Dissertation Abstracts International, Volume: 65-07, Section: B, page: 3767.

Thesis (Ph.D.)--The University of North Carolina at Chapel Hill, 2004.

This dissertation studies basic properties of the combinatorial landscape of Tversky's (1977) contrast model of similarity, and develops and validates a new algorithm (CONOPT) for fitting the contrast model to similarity data. Using methods from combinatorial landscape analysis, I show that the solution space for the feature recovery problem that algorithms for fitting the contrast model must solve is characterized by many local minima, with the number of minima growing as a power law in the size of the problem. Next, I present a new algorithm for fitting the contrast model. The algorithm recovers both theta weights and object features from similarity data, under the contrast model. An extensive set of numerical experiments using simulated data validates the algorithm. After validation, I apply the algorithm to several extant data sets taken from experiments on human similarity judgment. The algorithm recovers interpretable feature representations for some of the real data sets: for other data sets, the recovered representations are difficult to interpret. This latter observation reveals one of the greatest shortcomings of feature-based similarity models: Binary feature representations for large numbers of objects can be difficult to make sense of. Finally, general conclusions are drawn from this work.

ISBN: 049687425XSubjects--Topical Terms:

1017742
Psychology, Psychometrics.

An algorithm for fitting the contrast model.
LDR:02229nmm 2200289 4500 001 1846707
005 20051103093539.5
008 130614s2004 eng d
020 $a 049687425X
035 $a (UnM)AAI3140284
035 $a AAI3140284
040 $a UnM $c UnM
100 1 $a Bearden, J. Neil. $3 1934808
245 1 3 $a An algorithm for fitting the contrast model.
300 $a 68 p.
500 $a Source: Dissertation Abstracts International, Volume: 65-07, Section: B, page: 3767.
500 $a Director: Thomas S. Wallsten.
502 $a Thesis (Ph.D.)--The University of North Carolina at Chapel Hill, 2004.
520 $a This dissertation studies basic properties of the combinatorial landscape of Tversky's (1977) contrast model of similarity, and develops and validates a new algorithm (CONOPT) for fitting the contrast model to similarity data. Using methods from combinatorial landscape analysis, I show that the solution space for the feature recovery problem that algorithms for fitting the contrast model must solve is characterized by many local minima, with the number of minima growing as a power law in the size of the problem. Next, I present a new algorithm for fitting the contrast model. The algorithm recovers both theta weights and object features from similarity data, under the contrast model. An extensive set of numerical experiments using simulated data validates the algorithm. After validation, I apply the algorithm to several extant data sets taken from experiments on human similarity judgment. The algorithm recovers interpretable feature representations for some of the real data sets: for other data sets, the recovered representations are difficult to interpret. This latter observation reveals one of the greatest shortcomings of feature-based similarity models: Binary feature representations for large numbers of objects can be difficult to make sense of. Finally, general conclusions are drawn from this work.
590 $a School code: 0153.
650 4 $a Psychology, Psychometrics. $3 1017742
650 4 $a Psychology, Cognitive. $3 1017810
650 4 $a Computer Science. $3 626642
690 $a 0632
690 $a 0633
690 $a 0984
710 2 0 $a The University of North Carolina at Chapel Hill. $3 1017449
773 0 $t Dissertation Abstracts International $g 65-07B.
790 1 0 $a Wallsten, Thomas S., $e advisor
790 $a 0153
791 $a Ph.D.
792 $a 2004
856 4 0 $u http://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=3140284