Abstract
Human Activity Recognition is a machine learning task for the classification of human physical activities. Applications for that task have been extensively researched in recent literature, specially due to the benefits of improving quality of life. Since wearable technologies and smartphones have become more ubiquitous, a large amount of information about a person's life has become available. However, since each person has a unique way of performing physical activities, a Human Activity Recognition system needs to be adapted to the characteristics of a person in order to maintain or improve accuracy. Additionally, when smartphones devices are used to collect data, it is necessary to manage its limited resources, so the system can efficiently work for long periods of time. In this paper, we present a semi-supervised ensemble algorithm and an extensive study of the influence of hyperparameter configuration in classification accuracy. We also investigate how the classification accuracy is affected by the person and the activities performed. Experimental results show that it is possible to maintain classification accuracy by adjusting hyperparameters, like window size and window overlap, depending on the person and activity performed. These results motivate the development of a system able to automatically adapt hyperparameter settings for the activity performed by each person.

Abstract
This study intends to explore the phenomenon of intrapreneurship and the relevance of the innovation processes in mature organizations by conducting three case studies on traditional industry sectors such as the textile, metalworking, and electronic components sector. The aim is to explore the role of entrepreneurial teams in such companies, to understand the inhibiting factors of innovation, and to find the essential conditions for these organizations to continuously innovate and promote intrapreneurship within their organizations. The findings reveal that several factors constrain employee participation in this process, including lack of qualifications levels, limited resources, corporate bureaucracy, and reduced time and budget. In the conditions necessary to support the innovation process emerge the support given by the management team, the qualification of individuals and teams, and the existence of an organizational culture that promotes communication, collaboration, and tolerance to risk.

Abstract

Abstract
Although hardware is generally seen as the main culprit for a computer's energy usage, software too has a tremendous impact on the energy spent. Unfortunately, there is still not enough support for software developers so they can make their code more energy-aware. This paper proposes a technique to detect energy inefficient fragments in the source code of a software system. Test cases are executed to obtain energy consumption measurements, and a statistical method, based on spectrum-based fault localization, is introduced to relate energy consumption to the source code. The result of our technique is an energy ranking of source code fragments pointing developers to possible energy leaks in their code. This technique was implemented in the SPELL toolkit. Finally, in order to evaluate our technique, we conducted an empirical study where we asked participants to optimize the energy efficiency of a software system using our tool, while also having two other groups using no tool assistance and a profiler, respectively. We showed statistical evidence that developers using our technique were able to improve the energy efficiency by 43% on average, and even out performing a profiler for energy optimization.

Abstract
This study studies a double-surface sliding-mode observer (DS-SMO) for estimating the flux and speed of induction motors (IMs). The SMO equations are based on an IM model in the stationary reference frame. The DS-SMO is developed based on the equations of a single-surface SMO (SS-SMO) of IM. In DS-SMO method, the observer is designed through combining sliding variables produced by combining estimated fluxes of currents error. The speed is easily determined based on the pass of switching signal through a low-pass filter. Also, an optimal DS-SMO (ODS-SMO) is proposed to improve the transient condition by optimally tuning the observer parameters. To optimise these parameters, the particle swarm optimisation method is adopted. Moreover, an improved DS-SMO (IDS-SMO) is proposed to improve both transient and steady-state conditions, torque ripple and total harmonic distortion. Moreover, the proposed IDS-SMO has a stable performance under sudden load change and the low-speed region. Finally, the accuracy of the proposed ODS-SMO and IDS-SMO methods is substantiated through simulation and experimental results.

Abstract
We extend dynamic logic with binders (for state variables) by distinguishing between observable and silent transitions. This differentiation gives rise to two kinds of observational interpretations of the logic: abstractor and behavioural specifications. Abstractor specifications relax the standard model class semantics of a specification by considering its closure under weak bisimulation. Behavioural specifications, however, rely on a behavioural satisfaction relation which relaxes the interpretation of state variables and the satisfaction of modal formulas phi and [alpha]phi by abstracting from silent transitions. A formal relation between abstractor and behavioural specifications is provided which shows that both coincide semantically under mild conditions. For the proof we instantiate the previously introduced concept of a behaviour-abstractor framework to the case of dynamic logic with binders and silent transitions.