Olson, C.: Parallel Algorithms for Hierarchical Clustering. Parallel Computing 21, (1995) 1313-1325  Home/Search   Document Details and Download   Summary   Related Articles   Check

....volume of the cell can be used to determine the density of the cell. The computational complexity and the quality of clustering is heavily dependent on grid size and density threshold parameters. A survey of parallel algorithms for hierarchical clustering using distance based metrics is given in [Ols95] These are more theoretical PRAM algorithms. Recently, k means algorithm has been parallelized [DM99] but is limited, however, in its applicability, as it requires the user to specify k, the number of clusters, and also does not find clusters in subspaces. Clusters are unions of connected high ....

C.F. Olson. Parallel algorithms for hierarchical clustering. Parallel Computing, 21, 1995.

....an impact on the cluster formation, and may lead to di erent interpretations (see e.g. Gor81, Eve78, Kru77, Web99] In general the above clustering algorithm runs in O(n ) where n is the number of entities. However, for the single linkage or nearest neighbour method O(n ) can be achieved [Ols95]. Applied to the above data set, one obtains the dendrogram in Fig. 3, which has ben additionally enriched by a colour map. 2.3 Spring Embedding Before we explain how to use spring embedding as a non linear multi dimensional scaling technique, let us review spring embedding. Spring embedding ....

Clark F. Olson. Parallel algorithms for hierarchical clustering. Parallel Computing, 21(8):1313-1325, 1995.

....1,i,j x i,2,j ) # # (dist dendrogramj (x 1 , x 2 ) Basically, it uses the distance defined in the actual dendrograms. The merging algorithm is costly but the main advantage is that not all the data needs to be communicated to one processor. For a parallel implementation of SLINK see [60]. There is a large variety of data mining clustering techniques including BIRCH [74] and CURE [38] A further class of clustering algorithms is based on density considerations, both using parametric models [22] and nonparameteric approaches in DBSCAN [31] ....

C. Olson. Parallel algorithms for hierarchical clustering. Parallel Computing, 8:1313--1325, 1995.

....this limitation different devices are used to introduce some sparsity into the matrix. This can be done by omitting entries smaller than a certain threshold, by using only a certain subset of data representatives, or by keeping with each point only a certain number of its nearest neighbors. In [Ols95] nearest neighbor chains are used to reduce the memory limitation. A sparse matrix can be further processed to better represent intuitive closeness or connectivity concepts. Notice that the way we process original (dis)similarity matrix, and the way derived linkage metrics are constructed reflects ....

....the concept of distance between subsets. Such a derived proximity measure is called a linkage metric. The type of the linkage metric significantly affects hierarchical algorithms, since it reflects the particular concept of closeness and connectivity. Major inter cluster linkage metrics [Mur85] [Ols95] include single link, average link, and complete link. The underlying dissimilarity measure (e.g. distance) is computed for every pair of points with one point in the first set and another point in the second set. A specific operation such as minimum (single link) average (average link) or ....

[Article contains additional citation context not shown here]

. Olson, C. Parallel algorithms for hierarchical clustering. Parallel Computing, 21, 1313-1325, 1995.

....j i d = 1) where j i represents the new cluster produced by the merge of clusters i and j. So, the distance between a new cluster (combining the two previous clusters i and j) and the cluster k, is a function of the previous distances between the implicated clusters, with coefficients [3] that are functions of the cardinality of these clusters. On the other hand, very different algorithms can be given for the same hierarchical clustering methods. However, a general agglomerative algorithm for hierarchical clustering may be described informally as follows: 1. For each cluster ....

....algorithm provides a good approximate method for these cases, but the order in which the points are examined can change the final hierarchy. Much work has been done employing parallel algorithms and parallel computer networks, for hierarchical clustering using several intercluster distance metrics [3]. 3 A Relative Dissimilarity Measure The classic algorithms for agglomerative hierarchical clustering are based on metrics, which only consider the absolute distance between two clusters, merging the nearest pair of clusters in each stage, i.e. the pair of clusters with highest absolute ....

C. F. Olson, "Parallel Algorithms for Hierarchical Clustering", Technical Report UCB//CSD-93-786, University of California at Berkeley, January 1993.

....pursuing are: Automation of the instrumentation of code installed in the VAN for management pur 15 poses. 5] proposes such an approach in their specific programming language. Automation of the grouping of nodes and events for the purpose of aggregating information within the VAN [9][10]. Study in more depth event correlation in an active networking environment. 8 ....

C. Olson, "Parallel Algorithms for Hierarchical Clustering", Parallel Computing, Vol. 21, 1995.

....can provide the appropriate setting where to execute clustering algorithms for extracting knowledge from largescale data repositories. Recently there has been an increasing interest in parallel implementations of data clustering algorithms. Parallel approaches to clustering can be found in [8, 4, 9, 5, 10]. In this paper we consider a parallel clustering algorithm based on Bayesian classification for distributed memory multicomputers. We propose a parallel implementation of the AutoClass algorithm, called P AutoClass, and validate by experimental measurements the scalability of our parallelization ....

C.F. Olson. Parallel Algorithms for Hierarchical Clustering. Parallel Computing, 21:13131325, 1995.

....usually retains the underlying cluster structure. Summarizing algorithms are based on the fact that data points that are very close to each other can be merged and their summary information can be used for efficient clustering [5, 18] Partitioning is used in [8] and in parallel algorithms in [12] where data is partitioned and each partition is then clustered and later the results are consolidated to determine a global set of clusters. These techniques help in reducing the size of the data while trying to retain the original cluster structure but they still require to run the traditional ....

....is the way out. Type (1) algorithm of previous paragraph that uses priority queues is a stored matrix method while type (2) algorithm is more suitable as stored data if M is not large and if memory is not enough to store O(N 2 ) similarity matrix. Recent algorithms on hierarchical clustering [8, 10, 12] uses priority queues. The algorithms presented here improves both stored matrix and stored data algorithms by reducing their CPU time significantly, and furthermore it reduces the memory requirement substantially for stored matrix algorithm. 3 Proposed Algorithms In the previous section we ....

[Article contains additional citation context not shown here]

C. F. Olson. Parallel algorithms for hierarchical clustering. Parallel Computing, 21:1313--1325, 1995.

....operation can be applied to many common tasks such as unsupervised classification, segmentation and dissection. We are focusing here on one specific algorithm for clustering namely k h means clustering [1] The original version of the k h means algorithm was designed for numerical data [4] 6][5]. Our contribution in this paper is the development of a parallel version of the k h means algorithm. We present a realization of this algorithm on top of a distributed object store installed on a simple PC network. 2 The k h means Algorithm The k h means belongs to the class of the partitional ....

C.F. Olson. Parallel algorithms for hierarchical clustering. Parallel Computing, 21, 1995.

No context found.

Olson, C.: Parallel Algorithms for Hierarchical Clustering. Parallel Computing 21, (1995) 1313-1325

No context found.

C. Olson. Parallel algorithms for hierarchical clustering. Parallel Computing, 21(8):1313--1325, 1995.

No context found.

C. Olson. Parallel algorithms for hierarchical clustering.

No context found.

C.F. Olson. Parallel algorithms for hierarchical clustering. Par. Comp., 21:1313-1325, 1995.

No context found.

C. Olson. Parallel algorithms for hierarchical clustering. Parallel Computing, 21(8):1313--1325, 1995.

No context found.

C. Olson. Parallel algorithms for hierarchical clustering. Parallel Computing, 21(8):1313--1325, 1995.

No context found.

C. Olson. Parallel algorithms for hierarchical clustering. Parallel Computing, 8:1313--1325, 1995.

No context found.

Olson, C.: Parallel Algorithms for Hierarchical Clustering. Parallel Computing 21, (1995) 1313-1325

No context found.

C. F. Olson, "Parallel algorithms for hierarchical clustering," Parallel Computing, vol. 21, no. 8, pp. 1313--1325, 1995.

No context found.

C. F. Olson. Parallel algorithms for hierarchical clustering. Parallel Computing, 21(8):1313--1325, 1995.

No context found.

C.F. Olson, Parallel Algorithms for Hierarchical Clustering, Parallel Computing, 21:1313--1325, 1995.

No context found.

C. Olson. Parallel algorithms for hierarchical clustering. Parallel Computing, 8:1313-1325, 1995.

No context found.

C.F. Olson, Parallel Algorithms for Hierarchical Clustering, Parallel Computing, 21:1313- 1325, 1995.

No context found.

Olson, C.: Parallel Algorithms for Hierarchical Clustering. Parallel Computing 21, (1995) 1313-1325

No context found.

Olson, C.: Parallel Algorithms for Hierarchical Clustering. Parallel Computing 21, (1995) 1313-1325

No context found.

Clark F. Olson, Parallel Algorithms for Hierarchical Clustering, Technical Report, Computer Science Division, Univ. of California at Berkeley, Dec.,1993.

First 50 documents

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