Abstract
There has been significant recent progress in the integration of probabilistic reasoning with first order logic representations (SRL). So far, the learning algorithms developed for these models all learn from scratch, assuming an invariant background knowledge. As an alternative, theory revision techniques have been shown to perform well on a variety of machine learning problems. These techniques start from an approximate initial theory and apply modifications in places that performed badly in classification. In this work we describe the first revision system for SRL classification, PFORTE, which addresses two problems: all examples must be classified, and they must be classified well. PFORTE uses a two step-approach. The completeness component uses generalization operators to address failed proofs and the classification component addresses classification problems using generalization and specialization operators. Experimental results show significant benefits from using theory revision techniques compared to learning from scratch. © Springer-Verlag Berlin Heidelberg 2006.

Abstract
We present a case study in which sequence mining algorithms were applied to web access log data. The data are from a portal that is targeted for business users. In this portal, like in many others, content is described using a set of descriptors, such as keywords, category and type. We investigate whether representing content by the type rather than its identifier enables existing sequence mining methods to obtain interesting patterns. Rather than a more traditional approach based on measures such as support and confidence, we analyze results from an application perspective. This enables us to identify opportunities for improving and extending these methods.

Abstract
This paper summarizes efforts in understanding the possible application of Information Theoretic Learning Principles to Power Systems. It presents the application of Renyi's Entropy combined with Parzen windows as a measure of information content of the error distribution in model parameter estimation in supervised learning. It illustrates the concept with an application to the prediction of power generated in a wind park, made by Takagi-Sugeno Fuzzy Inference Systems, whose parameters are discovered with an EPSO-Evolutionary Particle Swarm Optimization algorithm.

Abstract
FC Portugal Rescue team is the result of a cooperation project between the Universities of Aveiro and Porto in Portugal. Following previous collaborations of these two Portuguese Universities in RoboCup simulation league and associated competitions, this project intends to fully adapt the coordination methodologies developed by FC Portugal simulated soccer team to the search and rescue scenario. After FC Portugal's first successful participation in RoboCup rescue simulation league (Osaka, 2005), the team aimed at introducing new high-level strategic analysis tools in the Rescue league and developed FCPx tool. This tool enables an easy comparison of different teams' strategies, and the use of learning methodologies for high-level strategic decisions in rescue. The tool also enabled us to evaluate our coordination methodologies, analyzing the results achieved in several rescue parameters, and conclude that they are very promising for the search and rescue domain. FC Portugal Rescue team results are very promising achieving first place in RoboCup European Championship 2006 (Dutch Open) held in Eindhoven.

Abstract

Abstract
The fabrication and characterization of long-period gratings (LPGs) in fiber tapers is presented alongside supporting theory. The devices possess a high sensitivity to the index of aqueous solutions due to an observed spectral bifurcation effect, yielding a limiting index resolution of +/- 8.5 x 10(-5) for solutions with an index in the range 1.330-1.335.