The efficiency of a linear discriminant function based on unclassified initial samples

Gànesalingam, S.; McLachlan, Geoffrey J.

doi:10.1093/biomet/65.3.658

Cited by 68 publications

(17 citation statements)

References 5 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Cooper and Freeman [22] were optimistic enough about unlabeled data so as to title their work as "On the asymptotic improvement in the outcome of supervised learning provided by additional nonsupervised learning" . Other early studies, such as [23][24][25] further strengthened the assertion that unlabeled data should be used whenever available. Castelli [26] and Ratsaby and Venkatesh [27] showed that unlabeled data are always asymptotically useful for classification.…”

Section: Learning a Classifier From Labeled And Unlabeled Traininmentioning

confidence: 95%

“…Castelli [26] and Ratsaby and Venkatesh [27] showed that unlabeled data are always asymptotically useful for classification. Krishnan and Nandy [19,20] extended the results of [25] to provide efficiency results for discriminant and logistic-normal models for samples that are labeled stochastically. It should be noted that such previous theoretical work makes the critical assumption that p(C, X)…”

Section: Learning a Classifier From Labeled And Unlabeled Traininmentioning

confidence: 99%

See 1 more Smart Citation

Semisupervised learning of classifiers: theory, algorithms, and their application to human-computer interaction

Cohen

Cozman

Sebe

et al. 2004

IEEE Trans. Pattern Anal. Machine Intell.

232

138

View full text Add to dashboard Cite

Automatic classification is one of the basic tasks required in any pattern recognition and human computer interaction application. In this paper we discuss training probabilistic classifiers with labeled and unlabeled data. We provide a new analysis that shows under what conditions unlabeled data can be used in learning to improve classification performance. We also show that if the conditions are violated, using unlabeled data can be detrimental to classification performance. We discuss the implications of this analysis to a specific type of probabilistic classifiers, Bayesian networks, and propose a new structure learning algorithm that can utilize unlabeled data to improve classification. Finally, we show how the resulting algorithms are successfully employed in two applications related to human-computer interaction and pattern recognition; facial expression recognition and face detection.

show abstract

Section: Learning a Classifier From Labeled And Unlabeled Traininmentioning

confidence: 95%

Section: Learning a Classifier From Labeled And Unlabeled Traininmentioning

confidence: 99%

Semisupervised learning of classifiers: theory, algorithms, and their application to human-computer interaction

Cohen

Cozman

Sebe

et al. 2004

IEEE Trans. Pattern Anal. Machine Intell.

232

138

View full text Add to dashboard Cite

show abstract

“…This is the approach recom mended in the statistical literature when classifying a population for which no training sample is available to determine the discriminant function [Bryant and Wil liamson, 1978;Ganesalingam and McLachlan, 1978]. As noted above, the discrimi nant analysis is related to maximum likeli hood estimation of the relationships.…”

Section: Discussionmentioning

confidence: 99%

Evaluating Pedigree Data

Gm¹,

Jw²,

Ab³

et al. 1983

Hum Hered

View full text Add to dashboard Cite

Pedigree data can be evaluated, and subsequently corrected, by analysis of the distribution of genetic markers, taking account of the possibility of mistyping. Using a model of pedigree error developed previously, we obtained the maximum likelihood estimates of error parameters in pedigree data from Tokelau. Posterior probabilities for the possible true relationships in each family are conditional on the putative relationships and the marker data are calculated using the parameter estimates. These probabilities are used as a basis for discriminating between pedigree error and genetic marker error in families where inconsistencies have been observed. When applied to the Tokelau data and compared with the results of retyping inconsistent families, these statistical procedures are able to discriminate between pedigree and marker error, with approximately 90% accuracy, for families with two or more offspring. The large proportion of inconsistencies inferred to be due to marker error (61 %) indicates the importance of discriminating between error sources when judging the reliability of putative relationship data. Application of our model of pedigree error has proved to be an efficient way of determining and subsequently correcting sources of error in extensive pedigree data collected in large surveys.

show abstract

“…To reveal latent population structure within the craniometric dataset and to produce the individual posterior probabilities, or coefficients of membership, needed to estimate admixture proportions and infer geographic ancestry in the absence of population identifiers and reference samples, this study exploits the unsupervised model-based clustering methods (MBC) of finite mixture analysis (FMA). Finite mixture models are powerful tools for probabilistic data analysis as they provide a principled, yet flexible, framework for the robust clustering of various distributions and at all levels of supervision (Ganesalingam and McLachlan, 1978;Banfield and Raftery, 1993;Schroeter et al, 1998;McLachlan and Peel, 2000;Peel and McLachlan, 2000). Here, it is assumed that the data is composed of a mixture of a finite number of underlying Gaussian probability distributions, with each component in the model corresponding directly some number of unobserved clusters or populations, k, which may be mutually exclusive or exhibit varying degrees of overlap Fraley and Raftery, 2002).…”

Section: Unsupervised Model-based Clusteringmentioning

confidence: 99%

Population inference from contemporaryAmerican craniometrics

Algee‐Hewitt

2016

American J Phys Anthropol

View full text Add to dashboard Cite

Population inference methods that allow for the model-bound estimation of admixture and ancestry proportions from craniometric data not only enable parallel-skeletal and genetic-analyses but they are also shown to be more informative than those methods that perform hard classifications using externally-imposed categories or seek to explain gross variation by low-dimensional projections. Am J Phys Anthropol 160:604-624, 2016. © 2016 Wiley Periodicals, Inc.

show abstract

The efficiency of a linear discriminant function based on unclassified initial samples

Cited by 68 publications

References 5 publications

Semisupervised learning of classifiers: theory, algorithms, and their application to human-computer interaction

Semisupervised learning of classifiers: theory, algorithms, and their application to human-computer interaction

Evaluating Pedigree Data

Population inference from contemporaryAmerican craniometrics

Contact Info

Product

Resources

About