T. Kohonen, Self-Organization and Associative Memory, 2nd ed. Berlin: Springer Verlag, 1988.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....elements of a digital library as sample observations of an information continuum. Phenomena exhibiting continuous, gradual variation are commonly referred to as fields. The most common information visualization technique implementing a field concept is the self organizing map (SOM) method [23]. It creates a regular tessellation using uniform area units, akin to raster elements used in digital imagery and GIS. SOMs indeed behave similarly to standard raster data models, compared to the vector like behavior of the object conceptualizations discussed in the previous sections. For example, ....

T. Kohonen, Self-Organizing Maps. Berlin: Springer-Verlag, 1995.

....in response to the input signals and the value supplied by the transfer function. Various neural network models, differing in their structural details, are described in the literature. Some popular techniques for object extraction employ Hopfield [8, 9, 10] Kohonen selforganizing feature map [11] and the adaptive resonance theory (ART) 12] There have been several attempts to fuse the merits of fuzzy set theory and artificial neural networks under the heading of neuro fuzzy computing for improving the performance of the decision making systems with regard to the problem of object ....

T. Kohonen. Self-Organization and Associative Memory. Berlin: Springer-Verlag, 1989.

.... segmenting pulmonary trees [12] and brain tissue [13] Various model based supervised classifiers have been tested in segmenting 3 D brain images in [14, 15] Each region is associated with a multivariate Gaussian mixture density in [15] 3 D modeling from range images using self organizing maps [16] was employed in [17] Radial Basis Functions (RBF s) were used for 3 D iterative image reconstruction from projection data in [18] and in 3 D shape from shading reconstruction [19] In this study we employ an unsupervised classification procedure by using an RBF network for modeling the 3 D ....

....of the ellipsoid parameters by using an approach based on moments corresponds to a classical statistical formulation. Their estimation can be performed by means of the k means clustering algorithm or its adaptive implementation represented by the Learning Vector Quantizer (LVQ) algorithm [16]. A classical training algorithm for RBF networks employs the LVQ algorithm for estimating the Gaussian function parameters [29] Variants of this algorithm have been derived in order to increase its efficiency in modeling data distributions [26, 30] An algorithm based on median estimation ....

[Article contains additional citation context not shown here]

T. Kohonen, Self Organization and Associative Memory. Berlin: Springer-Verlag, 1988.

....unsupervised learning (or clustering) are employed in several elds. Among them, we have speech compression [1] image compression [2] pattern recognition [3] and computer vision [4] Several approaches to clustering exist in literature, both of the fuzzy type [5] 6] and of the hard type [7] [8]. Moreover, both of these kinds of algorithms can be further subdivided in c means techniques [7] 5] and competitive learning techniques [8] 9] 6] Some authors [6] 10] say that fuzzy algorithms are less sensitive to initial conditions than hard ones. This is true if we consider the ....

.... recognition [3] and computer vision [4] Several approaches to clustering exist in literature, both of the fuzzy type [5] 6] and of the hard type [7] 8] Moreover, both of these kinds of algorithms can be further subdivided in c means techniques [7] 5] and competitive learning techniques [8], 9] 6] Some authors [6] 10] say that fuzzy algorithms are less sensitive to initial conditions than hard ones. This is true if we consider the Generalized Lloyd Algorithm (GLA) 7] a hard c means technique, known also as LBG (from the initials of its authors) In [11] and [12] we ....

T. Kohonen, Self organization and associative memory. Berlin: Springer Verlag, 3rd ed., 1989.

....the Linde Buzo Gray (LBG) algorithm [32] belongs to the rst group. Recent developments in Neural Networks (NNs) architectures resulted in several competitive learning algorithms [51] as, for example, the well known Learning Vector Quantization (LVQ) and the Self Organizing Feature Map (SOFM) [52]. Other competitive learning clustering techniques are Fuzzy LVQ (FLVQ) 49] Fuzzy Algorithms for LVQ (FALVQ) 53, 54] Generalized LVQ (GLVQ) 55] and GLVQ Fuzzy (GLVQ F) 56] The performance of several VQ algorithms depends on the choice of the initial conditions and the con guration ....

T. Kohonen, Self organization and associative memory. Berlin: Springer Verlag, 3rd ed., 1989.

....layer, and weights on connection between consequent layers. In order to avoid these tricky and difficult situations, self generating neu ral networks (SGNNs) are focussed an attention because of their simplicity on networks design. SGNNs are some kinds of extension of self organizing maps (SOMs) [7] and utilize the competitive learning algorithm which is implemented as self generating neural tree (SGNT) The SGNT algorithm is proposed in [14] to generate a neural tree automatically from a training data set directly. Originally, this SGNT algorithm is basically hierarchical clustering ....

T. Kohonen, Self-Organizing Maps, Berlin: Springer-Verlag, 1995.

....to this class. This method is not as eOEcient as standard PCA in approximating the input vectors x(t) but it implicitly takes into account higher order statistics in computations, and can be used for blind separation after prewhitening [13, 3, 1] Also the well known self organizing map (SOM) [14, 15, 16] can be expressed in terms of the model (6) In SOM, only one of the coeOEcients g j (x(t) W) is nonzero and actually equal to unity. This coeOEcient corresponds to the weight (basis) vector w l which is closest to the data vector x(t) in Euclidean norm. By learning the weight vectors w 1 ; ....

....in K means clustering, and the city block distance for super Gaussian densities. The experimental results in the next section show that this somewhat intuitive argument is indeed valid. 4. 3 Neural grouping methods The K means clustering algorithm is related with the self organizing map (SOM) [14, 15], in particular with the batch version of SOM. A major dioeerence between these two methods is that in SOM a neighborhood is used in forming the clusters. Usually the neighborhood is decreased during the learning process. See for example [14, 15, 16] for a detailed description of the SOM learning ....

[Article contains additional citation context not shown here]

T. Kohonen, Self-Organizing Maps. New York, Berlin: Springer-Verlag, 1995.

....found at level 2. The hierarchical structure is shown with a dark circle at root level, and subsequent levels represented by squares, triangles diamonds and open circles. 4 Other Neural Classifiers The most common and well known neural classifiers are Neural gas [8] and the Self Organising Map [7]. These two models were chosen for comparison with SCENT. Neural gas was proposed by Martinez et al. 8] The algorithm uses soft competition, in which many nodes move at each data presentation, where the degree of movement is based on the rank order of error. A temperature factor is also used ....

T. Kohonen, Self-Organizing Maps, 2 edn. Berlin: Springer-Verlag, 1997.

....unsupervised learning and supervised learning. Both types of ANN learning require a collection of training examples that enable the ANN to model the data set and produce accurate output values. Unsupervised learning systems; such as adaptive resonance theory (ART) 6] self organizing map (SOM) [24], or Hopfield [20] networks; do not require that the output value for a training sample be provided at the time of training, while supervised learning systems; such as backpropagation (multi layer perceptron) radial basis function (RBF) 35] counterpropagation [18] or fuzzy ARTMAP [7] a ....

....type (e.g. multiple linear or logistic) be specified in advance. However, regression ANNs suffer from the same constraints as regression models [14] such as the linear or curvilinear relationship of the data with heteroscedastic error [34] Likewise, learning vector quantization (LVQ) networks [24] try to divide input values into disjoint Heuristics Principles for the Design of Artificial Neural Networks Page 18 categories similar to discriminant analysis and consequently have the same data distribution requirements as discriminant analysis. The resource employee allocation example ....

Kohonen, T. Self-Organization and Associative Memory, Berlin: Springer-Verlag, 1988.

No context found.

, Self Organizing Map. Berlin: Springer-Verlag, 1995.

No context found.

, Self Organizing Map. Berlin: Springer-Verlag, 1995.

....to the visual quality of the image. B. Kohonen s Self Organizing Maps As mentioned in the introduction, the goal of this algorithm is to create a correspondence between the input space of stimuli and the output space constituted of the codebook elements, the codewords, or neurons. After learning [12], these last ones have to approximate the vectors in the input space in the best possible way. All neurons, or codewords, are physically arranged on a square grid; it is thus possible to define neighborhoods on the grid, which include all neurons whose distance (on the grid) from one (central) ....

T. Kohonen, Self-Organization and Associative Memory, 3rd ed. Berlin: Springer-Verlag, 1989.

No context found.

T. Kohonen, Self-Organization and Associative Memory, 2nd ed. Berlin: Springer Verlag, 1988.

No context found.

T. Kohonen, Self-Organizing Maps, Berlin: Springer-Verlag, 2nd edn., 1997.

No context found.

T. Kohonen, Self-Organizing Maps. Berlin: Springer-Verlag, 1995.

No context found.

T. Kohonen, Self-organization and associative memory, 2nd ed., Berlin: Springer Verlag, 1988.

No context found.

Kohonen T. Self-Organisation and Associative Memory. Berlin: Springer-Verlag, 1984

No context found.

T. Kohonen. Self-Organizing Maps. Berlin: Springer-Verlag, 1995.

No context found.

T. Kohonen, Self-Organization and Associative Memory (3rd ed.). Berlin: Springer-Verlag, 1989.

No context found.

T. Kohonen, Self-Organization and Associative Memory. Berlin: Springer-Verlag, 1984.

No context found.

T. Kohonen, Self-Organization and Associative Memory. Berlin: Springer-Verlag, 1984.

No context found.

T. Kohonen, Self-Organizing Maps. Berlin: Springer-Verlag, 1995.

No context found.

T. Kohonen, Self-Organization and Associative Memory (3rd ed.). Berlin: Springer-Verlag, 1989.

No context found.

T. Kohonen, Self-organization and associative memory, 2nd ed., Berlin: Springer Verlag, 1988.

No context found.

T. Kohonen, Self-Organizing Maps. Berlin: Springer-Verlag, 1995.

First 50 documents  Next 50

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC