Download now Free registration required
Clustering is a critical task in distributed exploratory and data mining scenarios whose goal is obtaining clusters of observations that best reveal underlying structures. Most often, wireless sensor networks offer considerable amounts of multidimensional observations rendering centralized clustering approaches impractical. This paper develops two decentralized k-means algorithms for clustering observations collected at spatially deployed wireless sensors. In both algorithms, sensors exchange sufficient information only with their one-hop neighbors. Surprisingly, simulations reveal that distributed k-means are less sensitive to initialization than their centralized counterparts resulting in superior performance with respect to the centralized k-means algorithm.
- Format: PDF
- Size: 148.6 KB