Multi-objective Evolutionary Clustering Algorithm Research Articles

Multitask Learning for Joint Diagnosis of Multiple Mental Disorders in Resting-State fMRI.

Facing the increasing worldwide prevalence of mental disorders, the symptom-based diagnostic criteria struggle to address the urgent public health concern due to the global shortfall in well-qualified professionals. Thanks to the recent advances in neuroimaging techniques, functional magnetic resonance imaging (fMRI) has surfaced as a new solution to characterize neuropathological biomarkers for detecting functional connectivity (FC) anomalies in mental disorders. However, the existing computer-aided diagnosis models for fMRI analysis suffer from unstable performance on large datasets. To address this issue, we propose an efficient multitask learning (MTL) framework for joint diagnosis of multiple mental disorders using resting-state fMRI data. A novel multiobjective evolutionary clustering algorithm is presented to group regions of interests (ROIs) into different clusters for FC pattern analysis. On the optimal clustering solution, the multicluster multigate mixture-of-expert model is used for the final classification by capturing the highly consistent feature patterns among related diagnostic tasks. Extensive simulation experiments demonstrate that the performance of the proposed framework is superior to that of the other state-of-the-art methods. Moreover, the potential for practical application of the framework is also validated in terms of limited computational resources, real-time analysis, and insufficient training data. The proposed model can identify the remarkable interpretative biomarkers associated with specific mental disorders for clinical interpretation analysis.

A knee point driven Kriging-assisted multi-objective robust fuzzy clustering algorithm for image segmentation

To enhance the segmentation performance and optimization efficiency of multi-objective evolutionary clustering algorithms for noisy images, this study presents a knee point driven Kriging-assisted multi-objective robust fuzzy clustering algorithm for image segmentation (K2 MORFC). First, pixel and region-level fitness functions are developed to comprehensively consider the spatial constraints and region consistency in an image. In the pixel-level fitness function, a Kullback–Leibler (KL) divergence-based spatial constraint term is employed and combined with intra-class compactness to preserve the membership similarity between an arbitrary image pixel and its non-local neighbors. In the region-level fitness function, image edge information is utilized to evaluate the regional consistency of the partition. Moreover, an adaptive determination mechanism of the weight factor for the spatial constraint term is designed in the pixel-level fitness function to control the influence of the spatial constraint for each pixel. Furthermore, considering that the preference toward knee points can effectively improve the convergence performance, a knee point driven Kriging-assisted evolutionary algorithm (K2EA) is proposed to efficiently optimize these two fitness functions, including a dynamic subspace knee point searching strategy, a knee point guided environmental selection strategy, and a Kriging model management mechanism. Finally, we construct a fuzzy clustering validity index with a KL divergence-based non-local spatial constraint term to select the optimal solution from the nondominated solution set. Competitive experimental results on the DTLZ benchmark verify the effectiveness of K2EA. Segmentation results on color images from the Berkeley dataset and magnetic resonance images from the BrainWeb and IBSR datasets confirm the performance of K2MORFC.

An Empirical Study of Cluster-Based MOEA/D Bare Bones PSO for Data Clustering †

Previously, cluster-based multi or many objective function techniques were proposed to reduce the Pareto set. Recently, researchers proposed such techniques to find better solutions in the objective space to solve engineering problems. In this work, we applied a cluster-based approach for solution selection in a multiobjective evolutionary algorithm based on decomposition with bare bones particle swarm optimization for data clustering and investigated its clustering performance. In our previous work, we found that MOEA/D with BBPSO performed the best on 10 datasets. Here, we extend this work applying a cluster-based approach tested on 13 UCI datasets. We compared with six multiobjective evolutionary clustering algorithms from the existing literature and ten from our previous work. The proposed technique was found to perform well on datasets highly overlapping clusters, such as CMC and Sonar. So far, we found only one work that used cluster-based MOEA for clustering data, the hierarchical topology multiobjective clustering algorithm. All other cluster-based MOEA found were used to solve other problems that are not data clustering problems. By clustering Pareto solutions and evaluating new candidates against the found cluster representatives, local search is introduced in the solution selection process within the objective space, which can be effective on datasets with highly overlapping clusters. This is an added layer of search control in the objective space. The results are found to be promising, prompting different areas of future research which are discussed, including the study of its effects with an increasing number of clusters as well as with other objective functions.

Coarse–fine surrogate model driven multiobjective evolutionary fuzzy clustering algorithm with dual memberships for noisy image segmentation

To improve the segmentation performance and boost evolutionary efficiency of multiobjective evolutionary clustering algorithms on noisy images, this paper proposes a coarse–fine surrogate model driven multiobjective evolutionary fuzzy clustering algorithm with dual membership functions (CFS-MOEFC). The existing fuzzy clustering validity-based fitness functions generally consider one single fuzzy membership function, which cannot fully process the uncertainty in an image. Moreover, most existing fitness functions used in image segmentation only utilize one type of spatial information derived from the image, which cannot behave robust on images with uncertain types of noise. To deal with the above problems, dual fuzzy membership functions are first designed by utilizing the local and non-local image spatial information. Then, the designed dual membership functions and the two kinds of spatial information are both used to construct spatial information-motivated fitness functions for image segmentation. To promote the time efficiency, the coarse–fine surrogate model is employed to assist the evolutionary optimization process, such as approximating the fitness functions instead of real function evaluations and evolving satisfactory cluster centers for CFS-MOEFC. Besides, a dual memberships-driven cluster validity index combining the local and non-local spatial information is designed for selecting an optimal solution from the final non-dominated solution set of CFS-MOEFC. Experiments on Berkeley and Magnetic Resonance (MR) images indicate that CFS-MOEFC greatly improves the segmentation accuracy on images with multiple types of noise, preserves more significant detailed image information, and behaves well on time cost comparing with state-of-the-art methods.

Multiobjective multiple features fusion: A case study in image segmentation

Most of existing image segmentation algorithms are only based on the color feature. However, the spatial distribution of an image can not be well described by using the color feature alone. Thus, it is necessary to add additional features to design efficient segmentation algorithms. Although many researchers also try to use multiple features for image segmentation, it is extremely difficult to combine multiple features automatically. This paper proposes a multiojective multiple features fusion strategy for image segmentation. The basic idea is to convert the segmentation problem into a multiobjective optimization problem, in which each objective considers one feature. It contains three steps. First, the original image is split into a set of over-segmented regions by using Meanshift to preserve the spatial details and to simplify the segmentation problem. Second, both the color and texture features are extracted to describe the regions. And two similarity matrices are designed by computing the similarity between each pair of regions in two features respectively. Third, a multiobjective evolutionary clustering algorithm is applied to merge these over-segmented regions. In this stage, two objective functions are designed based on the color and texture features respectively. A region index encoding scheme is introduced to design the individual, which contains some cluster representative regions. Some evolutionary operators are proposed to generate the new population. In the final generation, the best solution is selected from nondominated solutions for subsequent segmentation. Experiment results show that the proposed method provides promising segmentation results in combining the color and texture features.

Multi-objective evolutionary clustering with complex networks

Evolutionary clustering (EC) refers to the applications of evolutionary optimization algorithms such as genetic algorithm to data clustering. Although multi-objective evolutionary clustering algorithms were proposed to simultaneously consider different cluster properties such as compactness and separation, these techniques usually suffer from a reasonable initial population and a pre-defined number of clusters. Besides, the effectiveness of evolutionary operators is decreased in dealing with the clustering problem. On the other side, complex networks play an essential role in different fields of machine learning. In a complex network, points are considered as nodes, and the dataset is shown as a connected weighted graph. Also, complex networks tend to present a modular structure. This paper applies two concepts of complex networks including node centrality and community modularity to introduce a novel multi-objective evolutionary clustering. The proposed centrality modularity-based multi-objective evolutionary clustering (CMMOEC) takes the advantage of nodes similarity to find the best initial population of clustering solutions and provide new structural-based modularity to determine the optimal number of clusters automatically. Moreover, the proposed modularity is used to design a new recombination and mutation operator so that it generates offspring solutions that satisfy more diversity. Experiments carried out on several artificial and real-world datasets with different structures. The performance of the proposed algorithm is evaluated by the Adjusted Rand Index (ARI). Simulation results indicate that the proposed algorithm satisfies better performance in comparison to traditional methods.

Detecting the evolving community structure in dynamic social networks

Identifying the evolving community structure of social networks has recently drawn increasing attention. Evolutionary clustering, previously proposed to detect the evolution of clusters over time, presents a temporal smoothness framework to simultaneously maximize clustering accuracy and minimize the clustering drift between two successive time steps. Under this framework, evolving patterns of communities in dynamic networks were detected by finding the best trade-off between the clustering accuracy and temporal smoothness. However, two main drawbacks in previous methods limit the effectiveness of dynamic community detection. One is that the classic operators implemented by existing methods cannot avoid that a node is often inter-connected to most of its neighbors. The other is that those methods take it for granted that an inter-connection cannot exist between nodes clustered into the same community, which results in a limited search space. In this paper, we propose a novel multi-objective evolutionary clustering algorithm called DECS, to detect the evolving community structure in dynamic social networks. Specifically, we develop a migration operator cooperating with efficient operators to ensure that nodes and their most neighbors are grouped together, and use a genome matrix encoding the structure information of networks to expand the search space. DECS calculates the modularity based on the genome matrix as one of objectives to optimize. Experimental results on synthetic networks and real-world social networks demonstrate that DECS outperforms in both clustering accuracy and smoothness, contrasted with other state-of-the-art methods.

Scaling-up multiobjective evolutionary clustering algorithms using stratification

Multiobjective evolutionary clustering algorithms are based on the optimization of several objective functions that guide the search following a cycle based on evolutionary algorithms. Their capabilities allow them to find better solutions than with conventional clustering algorithms when more than one criterion is necessary to obtain understandable patterns from the data. However, these kind of techniques are expensive in terms of computational time and memory usage, and specific strategies are required to ensure their successful scalability when facing large-scale data sets. This work proposes the application of a data subset approach for scaling-up multiobjective clustering algorithms and it also analyzes the impact of three stratification methods. The experiments show that the use of the proposed data subset approach improves the performance of multiobjective evolutionary clustering algorithms without considerably penalizing the accuracy of the final clustering solution.

Large-Scale Experimental Evaluation of Cluster Representations for Multiobjective Evolutionary Clustering

Multiobjective evolutionary clustering algorithms are based on the optimization of several objective functions that guide the search following a cycle based on evolutionary algorithms. Their capabilities allow them to find better solutions than with conventional clustering algorithms if the suitable individual representation is selected. This paper provides a detailed analysis of the three most relevant and useful representations-prototype-based, label-based, and graph-based-through a wide set of synthetic data sets. Moreover, they are also compared to relevant conventional clustering algorithms. Experiments show that multiobjective evolutionary clustering is competitive with regard to other clustering algorithms. Furthermore, the best scenario for each representation is also presented.

SAR image segmentation based on quantum-inspired multiobjective evolutionary clustering algorithm

The segmentation task in the feature space of an image can be formulated as an optimization problem. Recent researches have demonstrated that the clustering techniques, using only one objective may not obtain suitable solution because the single objective function just can provide satisfactory result to one kind of corresponding data set. In this letter, a novel multiobjective clustering approach, named a quantum-inspired multiobjective evolutionary clustering algorithm (QMEC), is proposed to deal with the problem of image segmentation, where two objectives are simultaneously optimized. Based on the concepts and principles of quantum computing, the multi-state quantum bits are used to represent individuals and quantum rotation gate strategy is used to update the probabilistic individuals. The proposed algorithm can take advantage of the multiobjective optimization mechanism and the superposition of quantum states, and therefore it has a good population diversity and search capabilities. Due to a set of nondominated solutions in multiobjective clustering problems, a simple heuristic method is adopted to select a preferred solution from the final Pareto front and the results show that a good image segmentation result is selected. Experiments on one simulated synthetic aperture radar (SAR) image and two real SAR images have shown the superiority of the QMEC over three other known algorithms.

Automatic clustering algorithm based on multi-objective Immunized PSO to classify actions of 3D human models

Multi-objective clustering algorithms are preferred over its conventional single objective counterparts as they incorporate additional knowledge on properties of data in the from of objectives to extract the underlying clusters present in many datasets. Researchers have recently proposed some standardized multi-objective evolutionary clustering algorithms based on genetic operations, particle swarm optimization, clonal selection principles, differential evolution and simulated annealing, etc. In many cases it is observed that hybrid evolutionary algorithms provide improved performance compared to that of individual algorithm. In this paper an automatic clustering algorithm MOIMPSO (Multi-objective Immunized Particle Swarm Optimization) is proposed, which is based on a recently developed hybrid evolutionary algorithm Immunized PSO. The proposed algorithm provides suitable Pareto optimal archive for unsupervised problems by automatically evolving the cluster centers and simultaneously optimizing two objective functions. In addition the algorithm provides a single best solution from the Pareto optimal archive which mostly satisfy the users' requirement. Rigorous simulation studies on 11 benchmark datasets demonstrate the superior performance of the proposed algorithm compared to that of the standardized automatic clustering algorithms such as MOCK, MOPSO and MOCLONAL. An interesting application of the proposed algorithm has also been demonstrated to classify the normal and aggressive actions of 3D human models.