A study of the classification capabilities of neural networks using unsupervised learning: A comparison withK-means clustering

Balakrishnan, P.V.; Cooper, Martha; Jacob, Varghese S.; Lewis, Philip A.

doi:10.1007/bf02294390

Cited by 100 publications

(62 citation statements)

References 19 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…For instance, Murtagh (1995) noted that the interpretation of the resulting maps from the SOM method is often difficult without subsequent cluster analysis. Balakrishnan et al (1994) indicated that for solutions with a small number of groupings, the SOM seems to be less competitive than conventional clustering algorithms. More importantly, many studies have indicated that in general, the SOM is a stochastic process, with its output being sensitive to the algorithm implementation settings including the initialisation of reference vectors, order of data presentation, learning rates and neighbourhood width (e.g.…”

Section: Introductionmentioning

confidence: 99%

On two different objective procedures for classifying synoptic weather types over east Australia

Jiang

Cheung

Luo

et al. 2011

Intl Journal of Climatology

View full text Add to dashboard Cite

Synoptic weather typing depends on the method and its implementation. This paper demonstrates the utility of a 2-phase batch self-organising map procedure (CP2) in comparison with a procedure (CP1) consisting of T-mode principal component analysis followed by convergent K-means clustering, for classifying the synoptic weather types over east Australia. Four classifications were obtained for the 52-year NCEP/NCAR dataset, one from CP1 and three from CP2. These classifications were examined in terms of grouping quality, mean type maps, type frequencies, lifetime and transitions, and in relation to the Southern Oscillation Index (SOI).The results showed that the classifications derived from different procedures are inter-confirmative and capture a similar set of major synoptic situations influencing east Australia, each having counterparts in previous studies and conforming well to local synoptic experience. In particular, this study demonstrates that CP2 is a promising tool for the purpose of weather typing with a two-fold utility: CP2 can perform data projection (neighbourhood width > = 1) and provide a flexible means for visualising the broad distribution of the daily weather patterns in the dataset; and the procedure can also conduct cluster analysis (neighbourhood width = 0) and produce results equivalent to those from CP1. It was found that performing data clustering or data projection may affect the inference associated with type frequencies. While focusing on methodology, the analysis has also shown that the frequency of synoptic weather types fluctuates on the seasonal, interannual, and decadal scales. From the 1970s to 1995 there was a decline in the occurrence of a few weather types associated with an east-coast trough extending over New South Wales from the north, and a slight increase in the activity of some anticyclonic types. Significant correlations were also identified between SOI and frequencies of synoptic weather types on the annual and seasonal timescales.

show abstract

Section: Introductionmentioning

confidence: 99%

On two different objective procedures for classifying synoptic weather types over east Australia

Jiang

Cheung

Luo

et al. 2011

Intl Journal of Climatology

View full text Add to dashboard Cite

show abstract

“…Conclusions seem to be ambivalent as different authors point to different conclusions, and no definitive results have emerged. Some authors (Flexer 1999;Balakrishnan, Cooper et al 1994;Waller, Kaiser et al 1998) suggest that SOM performs equal or worst than statistical approaches, while other authors conclude the opposite (Openshaw and Openshaw 1997;Openshaw, Blake et al 1995).…”

Section: Introductionmentioning

confidence: 99%

Self-organizing Maps as Substitutes for K-Means Clustering

Bação¹,

Lobo²,

Paìnho³

2005

Lecture Notes in Computer Science

188

View full text Add to dashboard Cite

Abstract. One of the most widely used clustering techniques used in GISc problems is the k-means algorithm. One of the most important issues in the correct use of k-means is the initialization procedure that ultimately determines which part of the solution space will be searched. In this paper we briefly review different initialization procedures, and propose Kohonen's SelfOrganizing Maps as the most convenient method, given the proper training parameters. Furthermore, we show that in the final stages of its training procedure the Self-Organizing Map algorithms is rigorously the same as the kmeans algorithm. Thus we propose the use of Self-Organizing Maps as possible substitutes for the more classical k-means clustering algorithms.

show abstract

“…Another observation, made by Balakrishnan et al [40] is that Kohonen feature maps are similar to the statistical technique of k-means clustering, yet it has been our observation that many papers describing an SOFM do not compare the efficacy of it with another visualization technique applied to the same dataset.…”

Section: Self-organizing Feature Mapsmentioning

confidence: 96%

Perspectives and Prospects

Mareschal¹

2016

Introduction to Drug Disposition and Pharmacokinetics

View full text Add to dashboard Cite

In 1998, over 400 papers on artificial neural networks (ANNs) were published in the context of medicine, but why is there this interest in ANNs? And how do ANNs compare with traditional statistical methods? We propose some answers to these questions, and go on to consider the 'black box' issue. Finally, we briefly look at two directions in which ANNs are likely to develop, namely the use of Bayesian statistics and knowledgedata fusion.

show abstract

A study of the classification capabilities of neural networks using unsupervised learning: A comparison withK-means clustering

Abstract: K-means, Kohonen, Monte Carlo simulation, neural networks, nonhierarchical clustering,

Cited by 100 publications

References 19 publications

On two different objective procedures for classifying synoptic weather types over east Australia

On two different objective procedures for classifying synoptic weather types over east Australia

Self-organizing Maps as Substitutes for K-Means Clustering

Perspectives and Prospects

Contact Info

Product

Resources

About