1 resultado para complete-linkage
em Massachusetts Institute of Technology
Filtro por publicador
- Acceda, el repositorio institucional de la Universidad de Las Palmas de Gran Canaria. España (1)
- AMS Tesi di Dottorato - Alm@DL - Università di Bologna (4)
- AMS Tesi di Laurea - Alm@DL - Università di Bologna (1)
- Aquatic Commons (5)
- Archive of European Integration (13)
- Archivo Digital para la Docencia y la Investigación - Repositorio Institucional de la Universidad del País Vasco (2)
- Biblioteca Digital | Sistema Integrado de Documentación | UNCuyo - UNCUYO. UNIVERSIDAD NACIONAL DE CUYO. (1)
- Biblioteca Digital da Produção Intelectual da Universidade de São Paulo (27)
- Biblioteca Digital da Produção Intelectual da Universidade de São Paulo (BDPI/USP) (14)
- Biodiversity Heritage Library, United States (5)
- BORIS: Bern Open Repository and Information System - Berna - Suiça (97)
- Boston University Digital Common (1)
- Brock University, Canada (11)
- Bucknell University Digital Commons - Pensilvania - USA (7)
- CaltechTHESIS (2)
- Cambridge University Engineering Department Publications Database (21)
- CentAUR: Central Archive University of Reading - UK (40)
- Center for Jewish History Digital Collections (1)
- Chinese Academy of Sciences Institutional Repositories Grid Portal (70)
- Cochin University of Science & Technology (CUSAT), India (2)
- Collection Of Biostatistics Research Archive (1)
- Comissão Econômica para a América Latina e o Caribe (CEPAL) (7)
- CUNY Academic Works (2)
- Dalarna University College Electronic Archive (2)
- Deakin Research Online - Australia (57)
- DI-fusion - The institutional repository of Université Libre de Bruxelles (1)
- Digital Archives@Colby (2)
- DigitalCommons@The Texas Medical Center (20)
- DigitalCommons@University of Nebraska - Lincoln (5)
- Digitale Sammlungen - Goethe-Universität Frankfurt am Main (5)
- DRUM (Digital Repository at the University of Maryland) (2)
- Duke University (3)
- eResearch Archive - Queensland Department of Agriculture; Fisheries and Forestry (12)
- Greenwich Academic Literature Archive - UK (1)
- Helda - Digital Repository of University of Helsinki (3)
- Indian Institute of Science - Bangalore - Índia (45)
- Massachusetts Institute of Technology (1)
- Ministerio de Cultura, Spain (16)
- National Center for Biotechnology Information - NCBI (51)
- Plymouth Marine Science Electronic Archive (PlyMSEA) (5)
- Publishing Network for Geoscientific & Environmental Data (16)
- QUB Research Portal - Research Directory and Institutional Repository for Queen's University Belfast (78)
- Queensland University of Technology - ePrints Archive (118)
- Repositório Científico do Instituto Politécnico de Lisboa - Portugal (1)
- Repositório digital da Fundação Getúlio Vargas - FGV (4)
- Repositório Institucional UNESP - Universidade Estadual Paulista "Julio de Mesquita Filho" (130)
- SAPIENTIA - Universidade do Algarve - Portugal (1)
- School of Medicine, Washington University, United States (3)
- Universidad Autónoma de Nuevo León, Mexico (5)
- Universidad de Alicante (1)
- Universidad del Rosario, Colombia (5)
- Universidad Politécnica de Madrid (7)
- Universidade Federal do Pará (1)
- Universidade Técnica de Lisboa (1)
- Universitätsbibliothek Kassel, Universität Kassel, Germany (1)
- Université de Lausanne, Switzerland (7)
- Université de Montréal, Canada (4)
- University of Connecticut - USA (1)
- University of Michigan (40)
- University of Southampton, United Kingdom (3)
- Worcester Research and Publications - Worcester Research and Publications - UK (1)
Resumo:
An increasing number of parameter estimation tasks involve the use of at least two information sources, one complete but limited, the other abundant but incomplete. Standard algorithms such as EM (or em) used in this context are unfortunately not stable in the sense that they can lead to a dramatic loss of accuracy with the inclusion of incomplete observations. We provide a more controlled solution to this problem through differential equations that govern the evolution of locally optimal solutions (fixed points) as a function of the source weighting. This approach permits us to explicitly identify any critical (bifurcation) points leading to choices unsupported by the available complete data. The approach readily applies to any graphical model in O(n^3) time where n is the number of parameters. We use the naive Bayes model to illustrate these ideas and demonstrate the effectiveness of our approach in the context of text classification problems.