Exploiting Data Topology in Visualization and Clustering of Self-Organizing Maps

Taşdemir, Kadim; Merényi, Erzsébet

doi:10.1109/tnn.2008.2005409

Cited by 156 publications

(77 citation statements)

References 18 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…2, left, except for one, have at most 3 neighbors. We see this from the numbers of connections to Voronoi neighbors (pairs of BMUs and second BMUs formed by weights and their Voronoi neighbors) [17], shown in the order of the most to least connected, in Table 1. The connection strength (the number of data samples selecting a weight and its Voronoi neighbor as a pair of BMU and second BMU) between the one weight that has a fourth neighbor, and that fourth most connected neighbor is 1 (negligible).…”

Section: Two Modes Of the Som During Supervised Learningmentioning

confidence: 86%

An SOM-Hybrid Supervised Model for the Prediction of Underlying Physical Parameters from Near-Infrared Planetary Spectra

Zhang

Merényi

Grundy

et al. 2009

Advances in Self-Organizing Maps

View full text Add to dashboard Cite

Abstract. Near-Infrared reflectance spectra of planets can be used to infer surface parameters, sometimes with relevance to recent geologic history. Accurate prediction of parameters (such as composition, temperature, grain size, crystalline state, and dilution of one species within another) is often difficult because parameters manifest subtle but significant details in noisy spectral observations, because diverse parameters may produce similar spectral signatures, and because of the high dimensionality of the feature vectors (spectra). These challenges are often unmet by traditional inference methods. We retrieve two underlying causes of the spectral shapes, temperature and grain size, with an SOM-hybrid supervised neural prediction model. We achieve 83.0±2.7% and 100.0±0.0% prediction accuracy for temperature and grain size, respectively. The key to these high accuracies is the exploitation of an interesting antagonistic relationship between the nature of the physical parameters, and the learning mode of the SOM in the neural model.

show abstract

Section: Two Modes Of the Som During Supervised Learningmentioning

confidence: 86%

An SOM-Hybrid Supervised Model for the Prediction of Underlying Physical Parameters from Near-Infrared Planetary Spectra

Zhang

Merényi

Grundy

et al. 2009

Advances in Self-Organizing Maps

View full text Add to dashboard Cite

show abstract

“…The robustness of SOM clustering method could be associated with its characterized non-linear projection from the higher dimensional space of inputs to a low dimensional grid, which facilitates the discovery of hidden patterns in the input data (Kohonen and Honkela, 2007;Moghimidarzi et al, 2016). The SOM proved to be able to handle large datasets with outliers effectively (Shahreza et al, 2011;Oyana et al, 2012), and it has been applied successfully in complex structures (Tasdemir and Merényi, 2009).…”

Section: Self-organizing Mapsmentioning

confidence: 99%

Enhanced velocity based pore pressure prediction using lithofacies clustering: A case study from a reservoir with complex lithology in Dezful Embayment, SW Iran

Arabameri¹,

soleymani²,

Tokhmechi³

2018

Preprint

View full text Add to dashboard Cite

Velocity based pore pressure prediction methods are widely accepted as a routine technique in the petroleum industry. Despite recent improvements, still, literature suffer from inconsistencies and uncertainties mostly arise from velocity anomalies due to complex lithostratigraphic setting or presence of various formation fluids. The primary goal of this paper is to improve the accuracy and reliability of the conventional Bowers and Tau methods in a reservoir with complex lithology. Our proposed workflow aims to improve the accuracy of the estimations by clustering the input data based on specific petrophysical characteristics. We show since each major zones at the offset test wells have a distinct compaction trend, empirical constants in Bowers and Tau methods can be calibrated for each cluster rather than the whole stratigraphic column. The clustering task was done by statistical analyses of a suite of well logs and validated with core derived lithologies. To find the best clustering algorithm, we applied and compared five techniques namely, K-means, basic sequential algorithmic scheme, single, and complete linkage hierarchical. We found that the self-organizing map (SOM) method provides the best results by maximizing lithology likelihood within each cluster and improve the overall accuracy of the Bowers and Tau methods. This research also aims to provide a systematic comparison of the mentioned clustering algorithms based on their ability in distinguishing various lithofacies. We also try to minimize the user interference in the process of clustering multiple lithofacies and improve the reproducibility of the results and demonstrate the capability of the proposed method through a case study in a reservoir in the Southwest of Iran. Satisfactory results of this study offer a safe ground for implementation of the proposed method in other sedimentary basins.

show abstract

“…However, as more focus is placed on cooperation processes, it becomes more difficult to visualize class structure or class boundaries, since cooperation processes have roles to diminish discontinuity between neurons related to class boundaries. Though several methods have been developed to measure and extract discontinuity on the output space [30], [31], it is still difficult to extract clear class structure.…”

Section: B Information-theoretic Sommentioning

confidence: 99%

Double Competition for Information-Theoretic SOM

Kamimura¹

2014

ijarai

View full text Add to dashboard Cite

Abstract-In this paper, we propose a new type of informationtheoretic method for the self-organizing maps (SOM), taking into account competition between competitive (output) neurons as well as input neurons. The method is called "double competition", as it considers competition between outputs as well as input neurons. By increasing information in input neurons, we expect to obtain more detailed information on input patterns through the information-theoretic method. We applied the informationtheoretic methods to two well-known data sets from the machine learning database, namely, the glass and dermatology data sets. We found that the information-theoretic method with double competition explicitly separated the different classes. On the other hand, without considering input neurons, class boundaries could not be explicitly identified. In addition, without considering input neurons, quantization and topographic errors were inversely related. This means that when the quantization errors decreased, topographic errors inversely increased. However, with double competition, this inverse relation between quantization and topographic errors was neutralized. Experimental results show that by incorporating information in input neurons, class structure could be clearly identified without degrading the map quality to severely.

show abstract

Exploiting Data Topology in Visualization and Clustering of Self-Organizing Maps

Cited by 156 publications

References 18 publications

An SOM-Hybrid Supervised Model for the Prediction of Underlying Physical Parameters from Near-Infrared Planetary Spectra

An SOM-Hybrid Supervised Model for the Prediction of Underlying Physical Parameters from Near-Infrared Planetary Spectra

Enhanced velocity based pore pressure prediction using lithofacies clustering: A case study from a reservoir with complex lithology in Dezful Embayment, SW Iran

Double Competition for Information-Theoretic SOM

Contact Info

Product

Resources

About