Abstract
Digital image methodologies related with Melanoma has become in the past years a major support for differential diagnosis in skin cancer. Computer Aided Diagnosis (CAD) systems, encompassing image acquisition, artifact removal, detection and selection of features, highlight Machine Learning algorithms as a novel strategy towards a digital assisted diagnosis in Dermatology. Although the central role played by color in dermoscopic image assessment, Machine Learning algorithms mainly use texture and shape features, derived from gray level images, obtained from the true color images of the skin. Since the acquisition conditions are key for the color characterization and thus, central for the quantification of different colors in a dermoscopic image, this work presents a strategy for color normalization, joint with its use in the calculation of the number of colors of a dermoscopic image. This methodology will contribute to the uniformity in the use of color features extracted from different datasets in CAD systems (acquired by distinct dermoscopes) possibly presenting distinct illumination characteristics. This normalization proposal can also be applied as an image preprocessing step, aimed to achieve higher scores in the standard metrics in ML algorithms.

Abstract
Many stochastic search algorithms are designed to optimize a fixed objective function to learn a task, i.e., if the objective function changes slightly, for example, due to a change in the situation or context of the task, relearning is required to adapt to the new context. For instance, if we want to learn a kicking movement for a soccer robot, we have to relearn the movement for different ball locations. Such relearning is undesired as it is highly inefficient and many applications require a fast adaptation to a new context/situation. Therefore, we investigate contextual stochastic search algorithms that can learn multiple, similar tasks simultaneously. Current contextual stochastic search methods are based on policy search algorithms and suffer from premature convergence and the need for parameter tuning. In this paper, we extend the well known CMA-ES algorithm to the contextual setting and illustrate its performance on several contextual tasks. Our new algorithm, called contextual CMAES, leverages from contextual learning while it preserves all the features of standard CMA-ES such as stability, avoidance of premature convergence, step size control and a minimal amount of parameter tuning.

Abstract

Abstract
Responsive design allows software developers to build a Web page that can dynamically adapt to the size of the devices. This development philosophy enables the rendering of Web pages in a fast and optimized way, ensuring a good user experience on mobile devices, tablet and desktop. In the scope of this study, we intend to explore the main advantages and limitations associated with responsive Web design. We adopted a quantitative approach based on a questionnaire filled by 181 professionals in the industry that allowed us to identify the reasons that lead software developers to the adoption of the responsive design and also address the limitations felt by them. The results obtained indicate that offering a good user experience and increasing accessibility stands out as being the most important advantages. On the other hand, the main limitations include the compatibility with older Web browsers, the higher loading time and the difficulties in optimizing user experience. Finally, it was found that the perception of the advantages and limitations of responsive design is distinct for professionals with more professional experience in the field and for freelancer developers. © 2017, Fernando Almeida and José Monteiro.

Abstract
This paper presents a teaching/learning experiment on the use of MITAppI2 as a friendly tool in Automation courses. The goal was to assess if the up-to-date mobile applications can act as promoters in learning automation topics. The experiment took place in two Portuguese universities. The results achieved point towards a successful use of these tools in university classes. © Springer International Publishing Switzerland 2017.

Abstract
An unsupervised method for convolutional neural network (CNN) architecture design is proposed. The method relies on a variable neighborhood search-based approach for finding CNN architectures and hyperparameter values that improve classification performance. For this purpose, t-Distributed Stochastic Neighbor Embedding (t-SNE) is applied to effectively represent the solution space in 2D. Then, k-Means clustering divides this representation space having in account the relative distance between neighbors. The algorithm is tested in the CIFAR-10 image dataset. The obtained solution improves the CNN validation loss by over 15% and the respective accuracy by 5%. Moreover, the network shows higher predictive power and robustness, validating our method for the optimization of CNN design.