Applying emphasized soft targets for Gaussian Mixture Model based classification

Jelali, Soufiane El; Lyhyaoui, Abdelouahid; Figueiras-Vidal, Anı́bal R.

doi:10.1109/imcsit.2008.4747229

Cited by 3 publications

(2 citation statements)

References 23 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Although there are many forms of generating soft versions of decision targets, such as convolutional smoothing [21], it seems reasonable to keep the advantages that sample emphasis provides. This was the orientation of [22] and of our previous work [23], in which we showed the effectiveness of using an emphasized combination of the original targets and those provided by an auxiliary classifier when the architectures are the popular Multi-Layer Perceptrons (MLP), and [24], with preliminary results for Gaussian Mixture Models (GMM).…”

Section: Introductionmentioning

confidence: 62%

Designing Model Based Classifiers by Emphasizing Soft Targets

Jelali

Lyhyaoui²,

Figueiras-Vidal

2009

Fundamenta Informaticae

Self Cite

View full text Add to dashboard Cite

When training machine classifiers, to replace hard classification targets by emphasized soft versions of them helps to reduce the negative effects of using standard cost functions as approximations to misclassification rates. This emphasis has the same kind of effect as sample editing methods, that have proved to be effective for improving classifiers performance. In this paper, we explore the effectiveness of using emphasized soft targets with generative models, such as Gaussian Mixture Models (GMM), and Gaussian Processes (GP). The interest of using GMM is that they offer advantages such as an easy interpretation and straightforward possibilities to deal with missing values. With respect to GP, if we use soft targets, we do not need to resort to any complex approximation to get a Gaussian Process classifier and, simultaneously, we can obtain the advantages provided by the use of an emphasis. Simulation results support the usefulness of the proposed approach to get better performance and show a low sensitivity to design parameters selection.

show abstract

Section: Introductionmentioning

confidence: 62%

Designing Model Based Classifiers by Emphasizing Soft Targets

Jelali

Lyhyaoui²,

Figueiras-Vidal

2009

Fundamenta Informaticae

Self Cite

View full text Add to dashboard Cite

show abstract

“…I propose a solution based on an ensemble of Support Vector Machines (SVM) of the kind described in [5] [6], and on a voting mechanism based on the Gaussian Mixture Model described in [7]. Common approach based on ensemble of classifiers like boosting and bagging focus on preparing different training data set for each member of ensemble [8] [9].…”

Section: A Proposed Solutionmentioning

confidence: 99%

Utilizing an ensemble of SVMs with GMM voting-based mechanism in predicting dangerous seismic events in active coal mines

Podlodowski¹

2016

Annals of Computer Science and Information Systems

View full text Add to dashboard Cite

Abstract-This paper presents an application of a Gaussian Mixture Model-based voting mechanism for an ensemble of Support Vector Machines (SVMs) to the problem of predicting dangerous seismic events in active coal mines. The author proposes a method of preparing an ensemble of SVMs with different parameters and using the "wisdom of the crowd" for a classification problem. Experiments performed during the research showed an improvement in the quality of the classification after the mixture of Gaussian distributions was applied as votes distribution. The author also proposes a method of data selection for long sequences of measurement arranged chronologically with highly unbalanced occurrence of the positive class in the twoclass classification problem. Finally, using the proposed model to solve the problem defined by the organizers of AAIA'16 DM showed an increase in the stability of the ensemble classifier and an improvement in the quality of the classification problem solution.

show abstract

Eliciting Confidence for Improving Crowdsourced Audio Annotations

Méndez

Cartwright

Bello

et al. 2022

Proc. ACM Hum.-Comput. Interact.

View full text Add to dashboard Cite

In this work we explore confidence elicitation methods for crowdsourcing "soft" labels, e.g., probability estimates, to reduce the annotation costs for domains with ambiguous data. Machine learning research has shown that such "soft" labels are more informative and can reduce the data requirements when training supervised machine learning models. By reducing the number of required labels, we can reduce the costs of slow annotation processes such as audio annotation. In our experiments we evaluated three confidence elicitation methods: 1) "No Confidence" elicitation, 2) "Simple Confidence" elicitation, and 3) "Betting" mechanism for confidence elicitation, at both individual (i.e., per participant) and aggregate (i.e., crowd) levels. In addition, we evaluated the interaction between confidence elicitation methods, annotation types (binary, probability, and z-score derived probability), and "soft" versus "hard" (i.e., binarized) aggregate labels. Our results show that both confidence elicitation mechanisms result in higher annotation quality than the "No Confidence" mechanism for binary annotations at both participant and recording levels. In addition, when aggregating labels at the recording level, results indicate that we can achieve comparable results to those with 10-participant aggregate annotations using fewer annotators if we aggregate "soft" labels instead of "hard" labels. These results suggest that for binary audio annotation using a confidence elicitation mechanism and aggregating continuous labels we can obtain higher annotation quality, more informative labels, with quality differences more pronounced with fewer participants. Finally, we propose a way of integrating these confidence elicitation methods into a two-stage, multi-label annotation pipeline.

show abstract

Applying emphasized soft targets for Gaussian Mixture Model based classification

Cited by 3 publications

References 23 publications

Designing Model Based Classifiers by Emphasizing Soft Targets

Designing Model Based Classifiers by Emphasizing Soft Targets

Utilizing an ensemble of SVMs with GMM voting-based mechanism in predicting dangerous seismic events in active coal mines

Eliciting Confidence for Improving Crowdsourced Audio Annotations

Contact Info

Product

Resources

About