Abstract
To get the most out of powerful tools, expert knowledge is often required. Experts are the ones with the suitable knowledge to tune the tools' parameters. In this paper we assess several techniques which can automatically fine-tune ANN parameters. Those techniques include the use of GA and stratified sampling. The fine-tuning includes the choice of the best ANN structure and the best network biases and their weights. Empirical results achieved in experiments performed using nine heterogeneous data sets show that the use of the proposed Stratified Sampling technique is advantageous. © 2009 Springer Science+Business Media, LLC.

Abstract
Climatic change emphasize the importance of biodiversity maintenance, Suggesting that germplasm adapted to organic, low input, or conventional conditions is needed to face future demands. This Study presents: I - The two steps genesis of the synthetic maize population 'Fandango', A) 'NUTICA' creation: in 1975, Miguel Mota and Silas Pego, initiated a new type of polycross method involving 77 yellow elite inbred lines (dent and flint; 20% Portuguese and 80% North American germplasm) from the NUMI programme (NUcleo de melhoramento de Milho, Braga, Portugal). These inbreds were intermated in natural isolation and progenies submitted to intensive selection for both parents during continued cycles; B) From 'NUTICA' to 'Fandango': Tandango' was composed of all the crosses that resulted from a North Carolina Design I matting design (1 male crossed with 5 females) applied to 'NUTICA'. II - The diversity evolution of 'Fandango' under a Participatory Breeding project at the Portuguese Sousa Valley region (VASO) initiated in 1985 by Pego, with CIMMYT support. Morphological, fasciation expression, and yield trials were conducted in Portugal (3 locations, 3 years) and in the USA (4 locations, I year) using seeds obtained from five to seven cycles of mass selection (MS). The selection across cycles wits clone by the breeder (until cycle 5) and farmer (before cycle II in present). ANOVA and regression analysis on the rate of direct response to selection were performed when the assumption of normality was positively confirmed. Otherwise the non parametric Multivariate Adaptive Regression Splines (MARS) was performed. Response to mass selection in lowa showed significant decrease in yield, while in Portugal a significant increase for time of silking, plant and ear height, ear diameters 2, 37 4, kernel number, cot) diameters, and rachis was observed. At this location also a significant decrease was observed for thousand kernel weight and ear length. These results showed that mass selection were not effective for significant yield increase, except when considered Lousada with breeder selection (3.09% of gain per cycle per year). Some non-para metric methods (MARS, decision trees and random forests) were used to get insights on the causes that explain yield in Fandango. Kernel weight and ear weight were the most important traits, although row numbers, number of kernels per row, ear length, and ear diameter were also of some importance influencing 'Fandango' yield.

Abstract
We propose a two-layer scheme to control a set of vehicles moving in a formation. The first; layer, file trajectory controller, is a nonlinear controller since most vehicles are nonholonomic systems and require a nonlinear, even discontinuous, feedback to stabilize them. The trajectory controller, a model predictive controller, computes centrally a bang-bang control law and only a small set of parameters need to be transmitted to each vehicle at each iteration. The second layer, the formation controller, aims to compensate for small changes around a nominal trajectory maintaining the relative positions between vehicles. We argue that; the formation control call be, in most; cases, adequately carried out, by a linear model predictive controller accommodating input, and state constraints. This has the advantage that the control laws for each vehicle are simple piecewise affine feedback laws that, call be pre-computed off-line and implemented in a, distributed way in each vehicle. Although several optimization problems have to be solved, the control strategy proposed results in a simple and efficient; implementation where no optimization problem needs to be solved in real-time at each vehicle.

Abstract
We report on the development of a Genetic Algorithm (GA), which has been integrated into a Decision Support System to plan the best assignment of the weekly self-promotion space for a TV station. The problem addressed consists on deciding which shows to advertise and when such that the number of viewers, of an intended group or target, is maximized. The GA proposed incorporates a greedy heuristic to find good initial solutions. These solutions, as well as the solutions later obtained through the use of the GA, go then through a repair procedure. This is used with two objectives, which are addressed in turn. Firstly, it checks the solution feasibility and if unfeasible it is fixed by removing some shows. Secondly, it tries to improve the solution by adding some extra shows. Since the problem faced by the commercial TV station is too big and has too many features it cannot be solved exactly. Therefore, in order to test the quality of the solutions provided by the proposed GA we have randomly generated some smaller problem instances. For these problems we have obtained solutions on average within 1% of the optimal solution value.

Abstract
This work proposes an exercise-dependent real options model for the valuation and optimal harvest timing of a forestry investment in eucalyptus. Investment in eucalyptus is complex, as trees allow for two cuts without replantation and have a specific time and growth window in which they are suitable for industrial processing into paper pulp. Thus, path dependency in the cutting options is observed, as the moment of exercise of the first option determines the time interval inwhich the second option may be exercised. Therefore, the value of the second option depends on the history of the state variables rather than on its final value. In addition, the options to abandon the project or convert land to another use, are also considered. The option value is estimated by solving a stochastic dynamic programming model. Results are reported for a case study in the Portuguese eucalyptus forest, which show that price uncertainty postpones the optimal cutting decisions.Moreover, optimal harvesting policies deviate from current practice of forest managers and allow for considerable gains. © Springer Science+Business Media, LLC 2009.

Abstract
In this work we propose a multi-population genetic algorithm for tree-shaped network design problems using random keys. Recent literature on finding optimal spanning trees suggests the use of genetic algorithms. Furthermore, random keys encoding has been proved efficient at dealing with problems where the relative order of tasks is important. Here we propose to use random keys for encoding trees. The topology of these trees is restricted, since no path from the root vertex to any other vertex may have more than a pre-defined number of arcs. In addition, the problems under consideration also exhibit the characteristic of flows. Therefore, we want to find a minimum cost tree satisfying all demand vertices and the pre-defined number of arcs. The contributions of this paper are twofold: on one hand we address a new problem, which is an extension of the well known NP-hard hop-constrained MST problem since we also consider determining arc flows such that vertices requirements are met at minimum cost and the cost functions considered include a fixed cost component and a nonlinear flow routing component; on the other hand, we propose a new genetic algorithm to efficiently find solutions to this problem.