Pattern matching for intelligence organizations is a challenging problem. The data sets are large and noisy, and there is a flexible and constantly changing notion of what constitutes a match. We are developing the Link Analysis Workbench (LAW) to assist an expert user in the intelligence community in creating and maintaining patterns, matching those patterns against a large collection of relational data, and manipulating partial results. This paper describes two key facets of the LAW system: (1) a pattern-matching module based on a graph edit distance metric, and (2) a system architecture that supports the integration and tasking of multiple pattern matching modules based on their capabilities and the specific problem at hand.
A typical human ability is the recognition of patterns in the world around us. It constitutes the basis of each natural science: the laws of physics, the description of species in biology, or the analysis of human behavior; they are all based on seeing patterns. Also in daily life pattern recognition plays an important role: reading texts, identifying people, retrieving objects, or finding the way in a city. Once patterns are established, learned from some examples or from a teacher, we are able to classify new objects or phenomena into a class of known patterns. The study of automatic pattern recognition has two sides, one purely fundamentally scientific and one applied.
A central algorithm in production systems is the pattern match among rule predicates and current data. Systems like OPS5 and its various derivatives use the RETE algorithm for this function. This paper describes and analyses several augmentations of the basic RETE algorithm that are incorporated into an experimental production system, YES/OPS, which achieve significant improvement in efficiency and rule clarity.
Progress in spatial databases, such as spatial data structures (Gating 1994), spatial reasoning (Egenhofer 1991), computational geometry (Preparata and Shames, 1985), etc., paved the way for the study of knowledge discovery in spatial databases which aims at the extraction of implicit knowledge, spatial relations, or other patterns not explicitly stored in spatial databases (Koperski, Adhikary and Hen 1996). Generally speaking, a spatial pattern is a pattern showing the interaction of two or more spatial objects or space-depending attributes according to a particular spacing or set of arrangements (DeMers 2000). For instance, cities across nations are often clustered near lakes, oceans and streams. Actually such an arrangement reveals a spatial association, meaning that one spatial pattern is totally or partially related to some other spatial pattern. Furthermore, questions can be raised about the causes not only of single distributions but also of spatially correlated distributions of phenomena. For instance, we may explain that the tendency of cities to cluster near water bodies is driven by the need for sources of drinking water Copyright 2000, American Association for Artificial Intelligence (www.aaai.org).
Sanhes, Jérémy (Université de Nouvelle-Calédonie) | Flouvat, Frédéric (Université de Nouvelle-Calédonie) | Pasquier, Claude (Institute of Biology Valrose, CNRS ) | Selmaoui-Folcher, Nazha (Université de Nouvelle-Calédonie) | Boulicaut, Jean-François (Université de Lyon, CNRS, INSA de Lyon)
Directed acyclic graphs can be used across many application domains. In this paper, we study a new pattern domain for supporting their analysis. Therefore, we propose the pattern language of weighted paths, primitive constraints that enable to specify their relevancy (e.g., frequency and compactness constraints), and algorithms that can compute the specified collections. It leads to a condensed representation setting whose efficiency and scalability are empirically studied.