Nonhierarchical Methods

Methods that cluster data on the basis of a single scan of the data set are referred to as single-pass. A proximity threshold is typically used to decide 

Use the first compound in the data set to start the first cluster. 

Calculate the similarity, using some appropriate measure, between the next compound and all the existing clusters. If its similarity to the most similar existing cluster exceeds some threshold, assign it to that cluster otherwise use it to start a new cluster. 

Repeat step 2 until all compounds have been assigned. 

This method is simple to implement and very fast. The major drawback is that it is order dependent if the compounds are rearranged and scanned in a different order, then the resulting clusters can be different. 

---

Nonhierarchical methods such as k-means and EM need to be initialized with k seeds. This presupposes that kisa reasonable estimation of the number... 

The widespread application of the Jarvis-Patrick nonhierarchical method exists in part because of the influence of the publications by Willett et but also because of the availability of the efficient commercial... 

Nonhierarchical methods such as that described by Jarvis and Patrick (61) have been widely used for compound selection from large databases (62). The principle behind the Jarvis-Patrick method is to group together compounds that have a large number of nearest neighbors in common. However, the... 

Clustering algorithms can be classified into four major approaches hierarchical methods, partitioning-based methods, density-based methods, and grid-based methods. Here, we will focus on the hierarchical cluster approach because it is often used in the context of structure-activity analysis. Recent research has suggested that hierarchical methods perform better than the more commonly used nonhierarchical methods in separating known actives and inactives [41]. 

Bacchus Yang 92 Piece of data Constantly removing details to simplify the search space Yes Discussed the Downward refinement property (DRP) and showed how hierarchical problem solving techniques that lack this property has no advantage over nonhierarchical methods... 

Nonhierarchical methods generate individual partitions rather than complete hierarchies of partitions and hence tend to be far less demanding of computational resources. The most efficient nonhierarchical methods have time and space complexities of OfMNJ (at most) and OfNJ, respectively, where M is the number of clusters generated. That said, the inverse relationship between efficiency and effectiveness that we have noted previously applies here, and the most effective of the nonhierarchical methods tend to be comparable in efficiency to the hierarchical methods. 

---