Abstract
Naive Bayes is a well known and studied algorithm both in statistics and machine learning. Bayesian learning algorithms represent each concept with a single probabilistic summary. In this paper we present an iterative approach to naive Bayes. The iterative Bayes begins with the distribution tables built by the naive Bayes. Those tables are iteratively updated in order to improve the probability class distribution associated with each training example. Experimental evaluation of Iterative Bayes on 25 benchmark datasets shows consistent gains in accuracy. An interesting side effect of our algorithm is that it shows to be robust to attribute dependencies.

Abstract
This paper presents a technique to construct compilers expressed in a strict, purely functional setting. The compilers do not rely on any explicit data structures, like trees, stacks or queues, to efficiently perform the compilation task. They are constructed as a set of functions which are directly called by the parser. An abstract syntax tree is neither constructed nor traversed. Such deforestated compilers are automatically derived from an attribute grammar specification. Furthermore this technique can be used to efficiently implement any multiple traversal algorithm.

Abstract
A reliable navigation system is a key factor for the success of an operational mission with an AUV in a real scenario. In this paper, we address the main issues involved in the implementation of a long baseline (LBL) navigation system for a REMUS AUV. This system replaces both the original hardware and software of the vehicle with a simpler, faster, less expensive and more precise system, based on a Kalman filter. We also discuss the influence of transponder location in the overall performance of the LBL navigation, and present results obtained with this new system in operational missions.

Abstract
In this paper, we analyse the ability of Profibus fieldbus to cope with the real-time requirements of a Distributed Computer Control System (DCCS), where messages associated to discrete events must be made available within a maximum bound time. Our methodology is based on the knowledge of real-time traffic characteristics, setting the network parameters in order to cope with timing requirements. Since non-real-time traffic characteristics are usually unknown at the design stage, we consider an operational profile where, constraining non-real-time traffic at the application level we assure that real-time requirements are met. Copyright (C) 1998 IFAC.

Abstract
This paper describes the specification and design of a prototype of a low cost open system for the inspection of underwater structures based on a remotely operated underwater vehicle under the project IES, a 3 year long effort funded by the Portuguese R&D program Praxis XXI. Unlike commercial approaches, a modular open system characterised by the incorporation of an on-board computer allowing for advanced control capabilities is envisaged. The control console is based on a standard PC and the tether is used only for power delivery and to establish a simple communication channel. In this project, we use advanced hybrid control techniques for sophisticated semi-autonomous operation management and control. The control architecture reuses part of the one designed for the underwater vehicle Isurus operated by the Laboratory of Underwater Systems and Technologies of Porto University. The implementation is designed in order to allow for multiple sensor configurations specified as add-ins. This leads to a dynamic, scalable and flexible system that can be easily configured according to the user specifications.

Abstract