CISE Postdoctoral Rsch Associates in Experimental Computer Science: Incremental Nearest Neighbor Algorithms for Similarity Retrivai in Multimedia Databases and Medical Applications

[901636] samet，美国马里兰大学实验计算机科学博士后助理：多媒体数据库和医学应用中相似性检索的增量最近邻算法相似搜索在多媒体数据库和其他涉及复杂对象的数据库应用中是一个非常重要的操作。两个物体之间的相似性用距离度量表示。讨论了处理相似度搜索的两种现有方法。第一种方法是基于对低维几何空间的映射（利用诸如r树之类的数据结构），而第二种方法是直接根据距离对对象进行索引（利用诸如m树之类的数据结构）。对于这两种方法，正在研究新的最近邻搜索增量算法的发展。这些算法将具有按相似度顺序逐个报告结果的特性，尽可能少地花费精力来生成每个新的结果对象。增量方面在预先不知道所需邻居数量的情况下提供了显著的好处。该算法在多媒体数据库、医学应用等领域的应用正在研究中。

9901636Samet, HananUniversity of MarylandCISE Postdoctoral Associates in Experimental Computer Science: IncrementalNearest Neighbor Algorithms for Similarity Retrieval in MultimediaDatabases and Medical ApplicationsSimilarity search is a very important operation in multimedia databases and other database applications involving complex objects. The similarity between two objects is represented with a distance metric. Two existing approaches for handling similarity search are addressed. The first is based on mapping to low-dimensional geometric space (making use of data structures such as the R-tree), while the second directly indexes the objects based on distances (making use of data structures such as the M-tree). For both approaches, the development of new incremental algorithms for nearest neighbor search are being investigated. The algorithms will have the property of reporting their result one-by-one in order ofsimilarity, with as little effort as possible expended to produce each new result object. The incremental aspect provides significant benefits in situations when the number of desired neighbors is unknown in advance. The application of the algorithms to problems in fields such as multimedia databases, medical application, as well as others, are being investigated.