• Communications for Statistical Applications and Methods
• /
• v.10 no.2
• /
• pp.507-517
• /
• 2003

Self-organizing map (SOM) has been developed mainly by T. Kohonen and his colleagues as a unsupervised learning neural network. Because of its topological ordering property, SOM is known to be very useful in pattern recognition and text information retrieval areas. Recently, data miners use Kohonen´s mapping method frequently in exploratory analyses of large data sets. One problem facing SOM builder is that there exists no sensible criterion for evaluating goodness-of-fit of the map at hand. In this short communication, we propose valid evaluation procedures for the Kohonen SOM of any size. The methods can be used in selecting the best map among several candidates.

• The Korean Journal of Applied Statistics
• /
• v.16 no.2
• /
• pp.321-333
• /
• 2003

Self-organizing map (SOM), a unsupervised learning neural network, has been developed by T. Kohonen since 1980's. Main application areas were pattern recognition and text retrieval. Because of that, it has not been spread to statisticians until late. Recently, SOM's are frequently drawn in data mining fields. Kohonen's SOM, however, needs improvements to become a statistician's standard tool. First, there should be a good guideline as for the size of map. Second, an enhanced visualization mode is wanted. In this study, principal components self-organizing map (PC-SOM), a modification of Kohonen's SOM, is proposed to meet such needs. PC-SOM performs one-dimensional SOM during the first stage to decompose input units into node weights and residuals. At the second stage, another one-dimensional SOM is applied to the residuals of the first stage. Finally, by putting together two stages, one obtains two-dimensional SOM. Such procedure can be easily expanded to construct three or more dimensional maps. The number of grid lines along the second axis is determined automatically, once that of the first axis is given by the data analyst. Furthermore, PC-SOM provides easily interpretable map axes. Such merits of PC-SOM are demonstrated with well-known Fisher's iris data and a simulated data set.

• Proceedings of the Korean Statistical Society Conference
• /
• 2004.11a
• /
• pp.219-225
• /
• 2004

코호넨(T. Kohonen)의 자기조직화지도(Self-Organizing Map; SOM)은 저차원 그리드 공간에 고차원 다변량 자료를 축약하여 시각적으로 나타내는 비지도 학습법의 일종으로 최근 들어 통계 분석자들이 많은 관심을 가지고 있는 분야이다. 그러나 SOM은 개체공간의 연속형으로 표현되는 개체를 저차원 그리드공간에 승자노드에 비연속적으로 표현한다는 단점을 지니고 있다. 본 논문에서는 SOM을 통계적 목적으로 사용하기 위해 요구되는 그리드공간에 개체를 연속적으로 표현하는 방법들을 제안하고 활용 예를 제시하고자 한다

• The Korean Journal of Applied Statistics
• /
• v.15 no.2
• /
• pp.251-267
• /
• 2002

Self Organizing Maps(SOM) by Kohonen is very fast algorithm in neural networks. But it doesn't show sure rules of training results. In this paper, we introduce to Bayesian Learning for Self Organizing Maps(BLSOM) which combines self organizing maps with Bayesian learning. So it supports explanatory power of models and improves prediction. BLSOM has global optima anywhere but SOM has not. This is proved by experiment in this paper.

• Proceedings of the Korean Institute of Intelligent Systems Conference
• /
• 1997.11a
• /
• pp.180-184
• /
• 1997

This paper presents the fuzzy neural network which utilizes a fuzzified Kohonen learning uses a fuzzy membership value, a function of the iteration, and a intra-membership value instead of a learning rate. The IRIS data set if used to test the fuzzy neural network. The test result shows the performance of the fuzzy neural network depends on k and the vigilance parameter T.

• The Korean Journal of Applied Statistics
• /
• v.18 no.1
• /
• pp.83-98
• /
• 2005

Exploring distributional patterns of multivariate data is very essential in understanding the characteristics of given data set, as well as in building plausible models for the data. For that purpose, low-dimensional visualization methods have been developed by many researchers along various directions. As one of methods, Kohonen's SOM (Self-Organizing Map) is prominent. SOM compresses the volume of the data, yields abstraction from the data and offers visual display on low-dimensional grids. Although it is proven quite effective, it has one undesirable property: SOM's display is discrete. In this study, we propose two techniques for enhancing quality of SOM's display, so that SOM's display becomes continuous. The proposed methods are demonstrated in two numerical examples.

• The Transactions of the Korea Information Processing Society
• /
• v.7 no.7
• /
• pp.2108-2115
• /
• 2000

The conventional Boolean retrieval systems based on vector spae model can provide the results of retrieval fast, they can't reflect exactly user's retrieval purpose including semantic information. Consequently, the results of retrieval process are very different from those users expected. This fact forces users to waste much time for finding expected documents among retrieved documents. In his paper, we designed a bayesian SOM(Self-Organizing feature Maps) in combination with bayesian statistical method and Kohonen network as a kind of unsupervised learning, then perform classifying documents depending on the semantic similarity to user query in real time. If it is difficult to observe statistical characteristics as there are less than 30 documents for clustering, the number of documents must be increased to at least 50. Also, to give high rank to the documents which is most similar to user query semantically among generalized classifications for generalized clusters, we find the similarity by means of Kohonen centroid of each document classification and adjust the secondary rank depending on the similarity.

• Journal of the Korean Institute of Telematics and Electronics
• /
• v.26 no.5
• /
• pp.85-91
• /
• 1989

In this paper, we propose a neural net system for speech recognition, which is composed of two neural networks. Firstly the self-supervised BP(Back Propagation) network generates the distributed concept corresponding to the activity pattern in the hidden units. And then the self-organizing neural network forms a concept map which directly displays the similarity relations between concepts. By doing the above, the difficulty in learning the conventional BP network is solved and the weak side of BP falling into a pattern matcher is gone, while the strong point of generating the various internal representations is used. And we have obtained the concept map which is more orderly than the Kohonen's SOFM. The proposed neural net system needs not any special preprocessing and has a self-learning ability.

• Journal of the Korean Institute of Intelligent Systems
• /
• v.11 no.9
• /
• pp.849-854
• /
• 2001

GTM(Generative Topographic Mapping) model is a probabilistic version of the SOM(Self Organizing Maps) which was proposed by T. Kohonen. The GTM is modelled by latent or hidden variables of probability distribution of data. It is a unique characteristic not implemented in SOM model, and, therefore, it is possible with GTM to analyze data accurately, thereby overcoming the limits of SOM. In the present investigation we proposed a BGTM(Bayesian GTM) combined with Bayesian learning and GTM model that has a small mis-classification ratio. By combining fast calculation ability and probabilistic distribution of data of GTM with correct reasoning based on Bayesian model, the BGTM model provided improved results, compared with existing models.

• The Transactions of the Korea Information Processing Society
• /
• v.2 no.6
• /
• pp.995-1001
• /
• 1995

The structure of a typical pattern recognition consists a pre-processing, a feature extraction(algorithm) and classification or recognition. In classification, when widely varying patterns exist in same category, we need the clustering which organize the similar patterns. Clustering algorithm is two approaches. Firs, statistical approaches which are k-means, ISODATA algorithm. Second, neural network approach which is T. Kohonen's LVQ(Learning Vector Quantization). Nikhil R. Palet al proposed the GLVQ(Generalized LVQ, 1993). This paper suggest the efficient feature extraction methods of handwritten numerals in GLVQ clustering network. We use the handwritten numeral data from 21's authors(ie, 200 patterns) and compare the proportion of misclassified patterns for each feature extraction methods. As results, when we use the projection combination method, the classification ratio is 98.5%.