Name:PILP (GISELA)
Description:Parallel Inductive Logic Programming
Abstract:This application is intended to discover hidden data from relational databases. It uses a technique called Inductive Logic Programming (ILP), where, given a background knowledge, a set of positive examples, a set of negative examples, and a language bias, the objective is to generate first order rules that (almost) perfectly describes all positive examples and none of the negative examples. We have been working with several domains, when applying ILP: drug discovery, analysis of mammograms, link discovery, among others. These domains present very large databases and sets of examples.<BR/>ILP systems have been quite successful in extracting comprehensible models of relational data. Indeed, for over a decade, ILP systems have been used to construct predictive models for data drawn from diverse domains. These include the sciences, engineering, language processing, environment monitoring, and software analysis. In a nutshell, ILP systems repeatedly examine candidate clauses (the “search space’’) to find good rules. Ideally, the search will stop when the rules cover nearly all positive examples with only a few negative examples being covered. Unfortunately, the search space can grow very quickly in ILP applications. Several techniques have therefore been proposed to improve search efficiency. Such techniques include improving computation times at individual nodes, better representations of the search, sampling the search space, and parallelism. Parallelism can be obtained from very different alternative approaches, such as dividing the search tree, dividing the examples, or even through performing cross-validation in parallel. An intriguing alternative approach that can lead to better accuracy whilst taking advantage of parallelism is the use of ensembles. Ensembles are classifiers that combine the predictions of multiple classifiers to produce a single prediction. To some extent, an induced theory is an ensemble of clauses. We can go one step further and combine different theories to form a single ensemble. The mai
Created:2011-05-10
Last updated:2011-05-10
Software
Virtual Organizations
Software Appliances
Container Catalog
Sites/Resource Providers
Applications Database
