Multiobjective data clustering

Law, Max W. K.; Topchy, A.; Jain, Anil K.

doi:10.1109/cvpr.2004.1315194

Cited by 129 publications

(77 citation statements)

References 4 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…For data with different types of cluster structures, other objective functions may be more appropriate [12]. On the other hand, it is also suggested that stability, which reflects the variation in the clustering solutions under perturbations should be considered in developing clustering algorithms [13].…”

Section: B Scalarized Multiobjective Learningmentioning

confidence: 99%

Pareto-Based Multiobjective Machine Learning: An Overview and Case Studies

Jin

Sendhoff

2008

IEEE Trans. Syst., Man, Cybern. C

322

View full text Add to dashboard Cite

Abstract-Machine learning is inherently a multiobjective task. Traditionally, however, either only one of the objectives is adopted as the cost function or multiple objectives are aggregated to a scalar cost function. This can be mainly attributed to the fact that most conventional learning algorithms can only deal with a scalar cost function. Over the last decade, efforts on solving machine learning problems using the Pareto-based multiobjective optimization methodology have gained increasing impetus, particularly due to the great success of multiobjective optimization using evolutionary algorithms and other population-based stochastic search methods. It has been shown that Pareto-based multiobjective learning approaches are more powerful compared to learning algorithms with a scalar cost function in addressing various topics of machine learning, such as clustering, feature selection, improvement of generalization ability, knowledge extraction, and ensemble generation. One common benefit of the different multiobjective learning approaches is that a deeper insight into the learning problem can be gained by analyzing the Pareto front composed of multiple Pareto-optimal solutions. This paper provides an overview of the existing research on multiobjective machine learning, focusing on supervised learning. In addition, a number of case studies are provided to illustrate the major benefits of the Pareto-based approach to machine learning, e.g., how to identify interpretable models and models that can generalize on unseen data from the obtained Pareto-optimal solutions. Three approaches to Pareto-based multiobjective ensemble generation are compared and discussed in detail. Finally, potentially interesting topics in multiobjective machine learning are suggested.

show abstract

Section: B Scalarized Multiobjective Learningmentioning

confidence: 99%

Pareto-Based Multiobjective Machine Learning: An Overview and Case Studies

Jin

Sendhoff

2008

IEEE Trans. Syst., Man, Cybern. C

322

View full text Add to dashboard Cite

show abstract

“…However, they have some problems: one index is rarely equally well applicable to different types of dataset, i.e. one criterion may not fit well with data distribution when regions of the data contains clusters of diverse shapes [20]; moreover, many of these algorithms often fall into local optima and sometimes they can deliver results that are meaningless for data analysis (e.g. when is optimized only separation validity index it could yield one group with all genes).…”

Section: Introductionmentioning

confidence: 99%

“…In this case, we called Multi-Objective Clustering (MOC), and it corresponds to the simultaneous optimization of several validity index. This approach is a two-step process: (i) discovery of clusters by any clustering algorithm during the optimization of a quality index, and (ii) construction of an "optimal" set of solutions that correspond to various tradeoffs between different objectives [20]. The work of [25] uses a MOC approach incorporating a method called SiMM-TS (Significant Multi-class Membership -Two Stage).…”

Section: Introductionmentioning

confidence: 99%

A Bi-Objective Clustering Algorithm for Gene Expression Data

Parraga-Alava¹,

Inostroza-Ponta²

2017

CLEIej

View full text Add to dashboard Cite

Clustering algorithms are a common method for data analysis in many science field. They have become popular among biologists because of ease to discovery similar cellular functions in gene expression data. Most approaches consider the gene clustering as an optimization problem, where an ad-hoc cluster quality index is optimized which can be defined regarding gene expression data or biological information. However, these approaches may not be sufficient since they cannot guarantee to generate clusters with similar expression patterns and biological coherence. In this paper, we propose a bi-objective clustering algorithm to discover clusters of genes with high levels of co-expression and biological coherence. Our approach uses a multi-objective evolutionary algorithm (MOEA) that optimizes two index based on gene expression level and biological functional classes. The algorithm is tested on three real-life gene expression datasets. Results show that the proposed model yields gene clusters with higher levels of co-expression and biological coherence than traditional approaches.

show abstract

“…For example, Saha et al observe "evidently one reason for the difficulty of clustering is that for many data sets no unambiguous partitioning of the data exists, or can be established, even by humans" [1]. Moreover, matters are even worse for applications that seek of non-homogeneous groupings, because "virtually all existing clustering algorithms assume a homogeneous clustering criterion over the entire feature space … because its intrinsic criterion may not fit well with the data distribution in the entire feature space" [2]; furthermore, "by focusing on just one aspect of cluster quality, most clustering algorithms … are not robust to variations in cluster shape, size, dimensionality and other characteristics" [3]. Finally, in specific application domains, users seek for clusters with similar extrinsic characteristics; their definitions of "interestingness" of clusters are usually different from those used in typical clustering algorithms.…”

Section: Introductionmentioning

confidence: 99%

A Framework for Multi-Objective Clustering and Its Application to Co-Location Mining

Jiamthapthaksin

Eick

Vilalta

2009

Advanced Data Mining and Applications

View full text Add to dashboard Cite

Abstract. The goal of multi-objective clustering (MOC) is to decompose a dataset into similar groups maximizing multiple objectives in parallel. In this paper, we provide a methodology, architecture and algorithms that, based on a large set of objectives, derive interesting clusters regarding two or more of those objectives. The proposed architecture relies on clustering algorithms that support plug-in fitness functions and on multi-run clustering in which clustering algorithms are run multiple times maximizing different subsets of objectives that are captured in compound fitness functions. MOC provides search engine type capabilities to users, enabling them to query a large set of clusters with respect to different objectives and thresholds. We evaluate the proposed MOC framework in a case study that centers on spatial co-location mining; the goal is to identify regions in which high levels of Arsenic concentrations are co-located with high concentrations of other chemicals in the Texas water supply.

show abstract

Multiobjective data clustering

Cited by 129 publications

References 4 publications

Pareto-Based Multiobjective Machine Learning: An Overview and Case Studies

Pareto-Based Multiobjective Machine Learning: An Overview and Case Studies

A Bi-Objective Clustering Algorithm for Gene Expression Data

A Framework for Multi-Objective Clustering and Its Application to Co-Location Mining

Contact Info

Product

Resources

About