Multi-Label Classification via Adaptive Resonance Theory-Based Clustering

Masuyama, Naoki; Nojima, Yusuke; Loo, Chu Kiong; Ishibuchi, Hisao

doi:10.1109/tpami.2022.3230414

Cited by 12 publications

(6 citation statements)

References 62 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The time complexity for training CA (Algorithm 1) is O(ndK) [30], where n is the number of d-dimensional training instances, and K is the number of nodes. In addition, RVEA-CA generates a CA network on the M -dimensional objective space by using the union of the population and the archive set as O(N ) training instances.…”

Section: Computational Complexity 1) Algorithmmentioning

confidence: 99%

“…RVEA-CA has two hyperparameters, i.e., the ratio (probability) p mat for the mating selection and the interval λ for the learning process of CA. Among hyperparameters in the learning procedure of CA (see Algorithm 1), the similarity threshold V has a much larger impact on the performance of CA than the interval λ [30], and we introduce an adjusting procedure for the similarity threshold V into RVEA-CA. On the other hand, we used a fixed ratio p mat = 0.5 for the computational experiments resulted in Section V-D.…”

Section: ) Performance Analysis With Convergence and Diversitymentioning

confidence: 99%

See 1 more Smart Citation

Reference Vector Adaptation and Mating Selection Strategy via Adaptive Resonance Theory-Based Clustering for Many-Objective Optimization

Kinoshita,

Masuyama,

Liu

et al. 2023

IEEE Access

Self Cite

View full text Add to dashboard Cite

Decomposition-based multiobjective evolutionary algorithms (MOEAs) with clustering-based reference vector adaptation show high optimization performance for many-objective optimization problems (MaOPs). Especially, algorithms that employ a clustering algorithm with a topological structure (i.e., a network composed of nodes and edges) show superior optimization performance to other MOEAs for MaOPs with irregular Pareto optimal fronts (PFs). These algorithms, however, do not effectively utilize information of the topological structure in the search process. Moreover, the clustering algorithms typically used in conventional studies have limited clustering performance, inhibiting the ability to extract useful information for the search process. This paper proposes an adaptive reference vector-guided evolutionary algorithm using an adaptive resonance theory-based clustering with a topological structure. The proposed algorithm utilizes the information of the topological structure not only for reference vector adaptation but also for mating selection. The proposed algorithm is compared with eight state-of-the-art MOEAs on 78 test problems. Experimental results reveal the outstanding optimization performance of the proposed algorithm over the others on MaOPs with various properties.

show abstract

Section: Computational Complexity 1) Algorithmmentioning

confidence: 99%

Section: ) Performance Analysis With Convergence and Diversitymentioning

confidence: 99%

Reference Vector Adaptation and Mating Selection Strategy via Adaptive Resonance Theory-Based Clustering for Many-Objective Optimization

Kinoshita,

Masuyama,

Liu

et al. 2023

IEEE Access

Self Cite

View full text Add to dashboard Cite

show abstract

“…In general, a clustering algorithm can be applied to classification tasks by using a clustering result (e.g., cluster centroids, topological networks) as a classifier [16], [40], [41], [42]. Note that, in many cases, a clustering-based classification algorithm can inherit the continual learning ability from a clustering algorithm [14], [24], [43].…”

Section: Clustering-based Classification Algorithms Capable Of Contin...mentioning

confidence: 99%

“…However, similar to ARTbased clustering algorithms, ART-based classification algorithms also suffer from the difficulty of the specification of parameters for maintaining good classification performance. A small number of studies have focused on specification/adjustment methods for significantly data-dependent parameters [16], [48], [49], [50], [51]. However, these studies still have parameters that need to be adjusted for each dataset.…”

Section: Clustering-based Classification Algorithms Capable Of Contin...mentioning

confidence: 99%

“…Adaptive Resonance Theory (ART) [9], [10] is one of the most successful approaches for handling the plasticitystability dilemma while realizing continual learning. ARTbased topological clustering algorithms [11], [12], which use Correntropy-Induced Metric (CIM) [13] as a similarity measure, show faster and more stable self-organizing ability than other ART-based algorithms [9], [14], [15], [16]. However, these algorithms have two data-dependent parameters, namely a similarity threshold (i.e., a vigilance parameter) for a node learning process and an edge deletion threshold for a topology adjustment method, which have a significant impact on their clustering performance.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Adaptive Resonance Theory-Based Topological Clustering With a Divisive Hierarchical Structure Capable of Continual Learning

Masuyama¹,

Amako

Yuna

et al. 2022

IEEE Access

View full text Add to dashboard Cite

Adaptive Resonance Theory (ART) is considered as an effective approach for realizing continual learning thanks to its ability to handle the plasticity-stability dilemma. In general, however, the clustering performance of ART-based algorithms strongly depends on the specification of a similarity threshold, i.e., a vigilance parameter, which is data-dependent and specified by hand. This paper proposes an ART-based topological clustering algorithm with a mechanism that automatically estimates a similarity threshold from the distribution of data points. In addition, for improving information extraction performance, a divisive hierarchical clustering algorithm capable of continual learning is proposed by introducing a hierarchical structure to the proposed algorithm. Experimental results demonstrate that the proposed algorithm has high clustering performance comparable with recently-proposed state-of-the-art hierarchical clustering algorithms.

show abstract

A Hybrid Principal Label Space Transformation-Based Binary Relevance Support Vector Machine and Q-Learning Algorithm for Multi-label Classification

Ebrahimi

2024

Arab J Sci Eng

View full text Add to dashboard Cite

Multi-Label Classification via Adaptive Resonance Theory-Based Clustering

Cited by 12 publications

References 62 publications

Reference Vector Adaptation and Mating Selection Strategy via Adaptive Resonance Theory-Based Clustering for Many-Objective Optimization

Reference Vector Adaptation and Mating Selection Strategy via Adaptive Resonance Theory-Based Clustering for Many-Objective Optimization

Adaptive Resonance Theory-Based Topological Clustering With a Divisive Hierarchical Structure Capable of Continual Learning

A Hybrid Principal Label Space Transformation-Based Binary Relevance Support Vector Machine and Q-Learning Algorithm for Multi-label Classification

Contact Info

Product

Resources

About