Projection-Based Clustering through Self-Organization and Swarm Intelligence

Thrun, Michael C.

doi:10.1007/978-3-658-20540-9

Cited by 63 publications

(108 citation statements)

References 89 publications

Supporting

Mentioning

108

Contrasting

Order By: Relevance

“…The Databionic swarm (DBS) implements a swarm of agents interacting with one another and sensing their environment. DBS can adapt itself to structures of high-dimensional data such as natural clusters characterised by distance and density based structures in the data space (Thrun, 2018). The algorithm consists of three modules: the non-linear projection method Pswarm, the visualisation technique of a topographic map based on the generalised U-matrix and the clustering approach itself.…”

Section: Methodsmentioning

confidence: 99%

“…Pswarm is a swarm of intelligent agents called DataBots (Ultsch, 2000). It is a parameter-free focusing projection method of a polar swarm that exploits concepts of selforganisation and swarm intelligence (Thrun, 2018). During construction of this type of projection, which is called the learning phase and requires an annealing scheme, structure analysis shifts from global optimisation to local distance preservation (focusing).…”

Section: Methodsmentioning

confidence: 99%

“…Demartines & Hérault, 1995;Ultsch & Lötsch, 2017;Van der Maaten & Hinton, 2008), the annealing scheme are data-driven, and therefore, this method does not require any parameters. During learning, each DataBot moves across the grid or stays in its current position in the search for the most potent scent that means it searches for other agents carrying data with the most similar features to itself with a data-driven decreasing search radius (Thrun, 2018). The movement of every DataBot is modelled using an approach of game theory, and the radius decreases only if a Nash equilibrium is found (Nash, 1951).…”

Section: Methodsmentioning

confidence: 99%

“…Each clustering criterion imposes a certain structure on the data, and if the data happen to conform to the requirements of a particular criterion, the true clusters are recovere' (Jain & Dubes, 1988, p. 91). Here, the Databionic swarm (DBS) is used to find natural clusters without imposing a particular structure on the data contrary to conventional algorithms (Thrun, 2018). The purpose of this work is to show that the cluster structures found with DBS are meaningful, new and interesting, whereas in | 219 Thrun (2018) the cluster structures were investigated from a methodological point of view, e.g.…”

Section: Introductionmentioning

confidence: 99%

“…Here, the Databionic swarm (DBS) is used to find natural clusters without imposing a particular structure on the data contrary to conventional algorithms (Thrun, 2018). The purpose of this work is to show that the cluster structures found with DBS are meaningful, new and interesting, whereas in | 219 Thrun (2018) the cluster structures were investigated from a methodological point of view, e.g. described in (Behnisch & Ultsch, 2015).…”

Section: Introductionmentioning

confidence: 99%

See 4 more Smart Citations

Cluster Analysis of Per Capita Gross Domestic Products

Thrun¹

2019

EBER

View full text Add to dashboard Cite

A B S T R A C T Objective:The purpose of this article is to show the value of exploratory data analysis performed on the multivariate time series dataset of gross domestic products per capita (GDP) of 160 countries for the years 1970-2010. New knowledge can be derived by applying cluster analysis to the time series of GDP to show how patterns in GDP can be explained in a data-driven way. Research Design & Methods:Patterns characterised by distance and density based structures were found in a topographic map by using dynamic time warping distances with the Databionic swarm (DBS) 1 . The topographic map represents a 3D landscape of data structures. Looking at the topographic map, the number of clusters was derived. Then, a DBS clustering was performed and the quality of the clustering was verified. Findings: Two clusters are identified in the topographic map. The rules deduced from classification and regression tree (CART) show that the clusters are defined by an event occurring in 2001 at which time the world economy was experiencing its first synchronised global recession in a quarter-century. Geographically, the first cluster mostly of African and Asian countries and the second cluster consists mostly of European and American countries. Implications & Recommendations: DBS can be used even by non-professionals in the field of data mining and knowledge discovery. DBS is the first swarm-based clustering technique that shows emergent properties while exploiting concepts of swarm intelligence, self-organisation, and game theory.

show abstract

Section: Methodsmentioning

confidence: 99%

Section: Methodsmentioning

confidence: 99%

Section: Methodsmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations

Cluster Analysis of Per Capita Gross Domestic Products

Thrun¹

2019

EBER

View full text Add to dashboard Cite

show abstract

Identification of critical hemodilution by artificial intelligence in bone marrow assessed for minimal residual disease analysis in acute myeloid leukemia: The Cinderella method

et al. 2022

View full text Add to dashboard Cite

Minimal residual disease (MRD) detection is a strong predictor for survival and relapse in acute myeloid leukemia (AML). MRD can be either determined by molecular assessment strategies or via multiparameter flow cytometry. The degree of bone marrow (BM) dilution with peripheral blood (PB) increases with aspiration volume causing consecutive underestimation of the residual AML blast amount. In order to prevent false‐negative MRD results, we developed Cinderella, a simple automated method for one‐tube simultaneous measurement of hemodilution in BM samples and MRD level. The explainable artificial intelligence (XAI) Cinderella was trained and validated with the digital raw data of a flow cytometric “8‐color” AML‐MRD antibody panel in 126 BM and 23 PB samples from 35 patients. Cinderella predicted PB dilution with high accordance compared to the results of the Holdrinet formula (Pearson's correlation coefficient r = 0.94, R2 = 0.89, p < 0.001). Unlike conventional neuronal networks Cinderella calculated the distributions of 12 different cell populations that were assigned to true hematopoietic counterparts as a human in the loop (HIL) approach. Besides characteristic BM cells such as myelocytes and myeloid progenitor cells the XAI identified discriminating populations, which were not specific for BM or PB (e.g., T cell/NK cell subpopulations and CD45 negative cells) and considered their frequency differences. Thus, Cinderella represents a HIL‐XAI algorithm capable to calculate the degree of hemodilution in BM samples with an AML MRD immunophenotype panel. It is explicable, transparent, and paves a simple way to prevent false negative MRD reports.

show abstract