GRAPH INDEXING AND RETRIEVAL BASED ON GRAPH PROTOTYPES
Xavier Cortés Llosa · 2012
[ANGLES] Taking a query from a high number of data stored into a database, as fast as possible, is a recurrent problem in the field of computer sciences practically since its origins. At the existence of this problem, it’s necessary to add, moreover, the fact that actually databases contains data types of more diverse and unexpected character possible. Now we are not talking about originating databases which only contained sets of numbers or characters strings. (...) All that I want to make into the present work and I think that was achieved as far as possible, has been to develop and to present a methodology to carry out this process. The Metric Trees of prototypes are based on a well-known strategy, which is based on grouping the data stored in database at the smartest possible way. But also we has added the concept of a graph prototype. A structure that contains information of a set of instances represented by graphs, used until now for classification and recognition. In this thesis we have used graphs as representatives of elements that have to be queried in databases. Note that graphs have the capacity to represent complex objects, for this reason the number of graph databases is increasing. Due to in the literature appears different ways to build a prototype, the work presented here shows a comparative study between the main methods. Combining these two concepts, the Metric Tree and the graph prototype, we propose the construction of metric trees where the graph prototypes are routing nodes to help to decide the way to explore when we make a search in the tree. We have used Metric Trees to make classification and to find all instances that are lower than a maximum distance. (...)