D. Alahakoon, S. K. Halgamuge, and B. Srinivasan, Dynamic self organizing maps with controlled growth for knowledge discovery, IEEE Transactions on Neural Networks, vol. 11, pp. 601 -- 614, 2000.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....data characteristics, it remains far from trivial to deter mine the network architecture that provides satisfy ing results. Thus, it certainly is worth considering neural network models that determine the number and arrangement of units during their unsupervised training process. We refer to [4, 5, 6] for recently proposed models that are based on the SOM, yet al.. low for adaptation of the network architecture dur ing training. Second, hierarchical relations between the input data are not mirrored in a straight forward fashion. Such relations are rather shown within the same representation ....

....of the in put space, the input patterns are again organized in one single fiat OM. The shortcoming of having to define the size of the SOM in advance has been addressed by a num ber of different models. Consider, for example the Incremental Grid Growing [4] Growing Grid [5] Growing OM [6], and the Hypercubical OM [24] The first, i.e. Incremental Grid Growing, allows the addition of new units at the boundary of the map. Furthermore, connections between units of the map may be established and removed according to some threshold settings based on the similarity of their respective ....

D. Alahakoon, S. K. Halgamuge, and B. Srini- vasan, "Dynamic self-organizing maps with controlled growth for knowledge discovery," IEEE Transactions on Neural Networks, vol. 11, no. 3, pp. 601-614, May 2000.

....input data characteristics, it remains far from trivial to determine the network architecture that provides satisfying results. Thus, it certainly is worth considering neural network models that determine the number and arrangement of units during their unsupervised training process. We refer to [4, 5, 6] for recently proposed models that are based on the SOM , yet al..low for adaptation of the network architecture during training. Second, hierarchical relations between the input data are not mirrored in a straight forward fashion. Such relations are rather shown within the same representation ....

....of the input space, the input patterns are again organized in one single at SOM . The shortcoming of having to de ne the size of the SOM in advance has been addressed by a number of di erent models. Consider, for example the Incremental Grid Growing [4] Growing Grid [5] Growing SOM [6], and the Hypercubical SOM [24] The rst, i.e. Incremental Grid Growing, allows the addition of new units at the boundary of the map. Furthermore, connections between units of the map may be established and removed according to some threshold settings based on the similarity of their respective ....

D. Alahakoon, S. K. Halgamuge, and B. Srinivasan, \Dynamic self-organizing maps with controlled growth for knowledge discovery," IEEE Transactions on Neural Networks, vol. 11, no. 3, pp. 601-614, May 2000.

....The more common model functions in current data mining practice include the following. 1) Classification [18] 22] classifies a data item into one of several predefined categorical classes. 2) Regression [8] 23] 25] maps a data item to a realvalued prediction variable. 3) Clustering [26] [33]: maps a data item into one of several clusters, where clusters are natural groupings of data items based on similarity metrics or probability density models. 4) Rule generation [34] 41] extracts classification rules from the data. 5) Discovering association rules [42] 45] describes ....

....with more than one million nodes to partition a little less than seven million patent abstracts where the documents are represented by 500 dimensional feature vectors. Vesanto et al. 32] employ a step wise strategy by partitioning the data with a SOM, followed by its clustering. Alahakoon et al. [33] perform hierarchical clustering of SOMs, based on a spread factor which is independent of the dimensionality of the data. Shalvi and DeClaris [29] have designed a data mining technique, combining Kohonen s self organizing neural network with data visualization, for clustering a set of ....

D. Alahakoon, S. K. Halgamuge, and B. Srinivasan, "Dynamic self organizing maps with controlled growth for knowledge discovery," IEEE Trans. Neural Networks, vol. 11, pp. 601--614, 2000.

....new units gradually using constructive techniques. This techniques can be classified as hierarchical and non hierarchical. Among the existing non hierarchical constructive algorithms for SOM are Growing Cell Structures [8] Growing Neural Gas (GNG) 9] and the Growing Self Organizing Map (GSOM) [1]. GNG starts the training process with two unit. The GNG starts the training process with two disconnected units and creates new units each patterns presented ( is a parameter) Each time a new unit is inserted, its weighting vector and error are obtained through of interpolation ....

D. Alahakoon, S.K. Halgamuge, and B. Srinivasan. Dynamic selforganizing map with controlled growth for knowledge dsicovery. IEEE Transaction, 11(3):601--614, 2000.

....error on a map unit increases a specified threshold. In the following the approach which is used in the presented application is briefly described. 2.1. A growing self organizing map approach The proposed method is mainly motivated by the growing self organizing map models presented in [1, 4]. In contrast to these approaches we use hexagonal map structure and restrict the algorithm to add new units to the external units if the accumulated error of a unit exceeds a specified threshold value. The algorithm can be briefly described as follows: 1. Predefine the initial grid size ....

D. Alahakoon, S. K. Halgamuge, and B. Srinivasan, Dynamic Self-Organizing Maps with Controlled Growth for Knowledge Discovery, IEEE Transactions on Neural Networks, 11(3), pp. 601614, 2000.

....It is trained over the whole training set. The dynamic growth is based on the criterion of neuron ambiguity (i.e. uncertainty about class assignment) which is quantified with an entropy measure that is defined over the CP SOM nodes. This differs from the local quantization error approach of [35] that grows the map at the nodes that accumulate the largest quantization error. We developed the entropy based growing technique because it accounts class ambiguity much better than the accumulated local error. The local error depends on the SOM quantization performance and thus can be large even ....

.... at the neighborhood of each ambiguous neuron a number of neurons that depends on its fuzziness (i.e. the more uncertain is the class of a neuron the more neurons are inserted over its neighborhood) The neuron insertion process is not described in detail since it follows the guidelines of [35]. 3.4.1.b Repeat the adaptation phase after the dynamic extension of the map in order to adjust the new neurons at the appropriate positions at the lattice and re execute the previous steps of the expansion phase reentering to test the conditions of 3.4 elseif NumTrainingSetAtAmbiguous ....

D. Alahakoon, S. K. Halgamuge, and B. Srinivasan, "Dynamic self-organizing maps with controlled growth for knowledge discovery," IEEE Trans. Neural Networks, vol. 11, pp. 601--614, May 2000. PAPADIMITRIOU et al.: ISCHEMIA DETECTION WITH AN SOM 515

....mining practice include: 1. Classification [18] 19] 20] 21] 22] classifies a data item into one of several predefined categorical classes. 2. Regression [8] 23] 24] 25] maps a data item to a real valued prediction variable. 3. Clustering [26] 27] 28] 29] 30] 31] 32] [33]: maps a data item into one of several clusters, where clusters are natural groupings of data items based on similarity metrics or probability density models. 4. Rule generation [34] 35] 36] 37] 38] 39] 40] 41] extracts classification rules from the data. 5. Discovering ....

....with more than one million nodes to partition a little less than seven million patent abstracts where the documents are represented by 500 dimensional feature vectors. Vesanto et al. 32] employ a step wise strategy by partitioning the data with a SOM, followed by its clustering. Alahakoon et al. [33] perform hierarchical clustering of SOMs, based on a spread factor which is independent of the dimensionality of the data. Shalvi and DeClaris [29] have designed a data mining technique, combining Kohonen s self organizing neural network with data visualization, for clustering a set of ....

D. Alahakoon, S. K. Halgamuge, and B. Srinivasan, "Dynamic self organizing maps with controlled growth for knowledge discovery," IEEE Transactions on Neural Networks, vol. 11, pp. 601--614, 2000.

....using a tree based organization of units. However, it does not focus on providing a hierarchical organization of data as all data are organized on one single at map. Various approaches address the problem of having to de ne the size of a SOM in advance, such as Dynamic Self Organizing Maps [1], Incremental Grid Growing [2] or Growing Grid [4] where new units are added to map areas where the data, i.e. the topics, are not represented at a satisfying degree of granularity. However, since all of these methods rely on the creation of a single at map, they tend to produce huge maps ....

D. Alahakoon, S. K. Halgamuge, and B. Srinivasan. Dynamic self-organizing maps with controlled growth for knowledge discovery. IEEE Trans Neural Networks, 11(3), 2000.

No context found.

D. Alahakoon, S. K. Halgamuge, and B. Srinivasan, Dynamic self organizing maps with controlled growth for knowledge discovery, IEEE Transactions on Neural Networks, vol. 11, pp. 601 -- 614, 2000.

No context found.

Alahakoon, D., Halgamuge, S.K., Srinivasan, B., Dynamic self organizing maps with controlled growth for knowledge discovery, IEEE Transactions on Neural Networks, vol. 11, pp.601-614, 2000.

No context found.

D. Alahakoon, S. H. Halgamuge, and B. Srinivasan, "Dynamic self-organizing maps with controlled growth for knowledge discovery", IEEE Trans Neural Networks, 11(3), 2000.

No context found.

Damminda Alahakoon, Saman K. Halgamuge, and Bala Srinivasan. Dynamic selforganizing maps with controlled growth for knowledge discovery. IEEE Transactions on Neural Networks, 11(3):601--614, May 2000.

No context found.

D. Alahakoon, S. K. Halgamuge, and B. Srinivasan. Dynamic self-organizing maps with controlled growth for knowledge discovery. IEEE Transactions on Neural Networks, 11(3), 2000.

No context found.

Alahakoon, D., Halgamuge, S. K., and Srinivasan, B. (2000). Dynamic self-organizing maps with controlled growth for knowledge discovery. IEEE Transactions on Neural Networks, 11(3):601--614.

No context found.

D. Alahakoon, S.K. Halgamuge and B. Srinivasan, "Dynamic self-organizing maps with controlled growth for knowledge discovery", IEEE Tractions on Neural Networks, vol. 11, no. 3, 2000, pp. 601-614.

No context found.

Alahakoon, D., Halgamuge, S.K., and Srinivasan, B. Dynamic self-organizing maps with controlled growth for knowledge discovery. IEEE Tractions on Neural Networks, 2000, 11(3):601-614

No context found.

D. Alahakoon, S. K. Halgamuge, and B. Srinivasan. Dynamic self-organizing maps with controlled growth for knowledge discovery. IEEE Transactions on Neural Networks, 11(3):601--614, 2000.

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