Abstract
CMA-ES is one of the most popular stochastic search algorithms. It performs favourably in many tasks without the need of extensive parameter tuning. The algorithm has many beneficial properties, including automatic step-size adaptation, efficient covariance updates that incorporates the current samples as well as the evolution path and its invariance properties. Its update rules are composed of well established heuristics where the theoretical foundations of some of these rules are also well understood. In this paper we will fully derive all CMA-ES update rules within the framework of expectation-maximisation-based stochastic search algorithms using information-geometric trust regions. We show that the use of the trust region results in similar updates to CMA-ES for the mean and the covariance matrix while it allows for the derivation of an improved update rule for the step-size. Our new algorithm, Trust-Region Co-variance Matrix Adaptation Evolution Strategy (TR-CMA-ES) is fully derived from first order optimization principles and performs favourably in compare to standard CMA-ES algorithm. © 2017 ACM.

Abstract
Stochastic search algorithms are black-box optimizers of an objective function. They have recently gained a lot of attention in operations research, machine learning and policy search of robot motor skills due to their ease of use and their generality. However, with slightly different tasks or objective functions, many stochastic search algorithms require complete re-learning in order to adapt the solution to the new objective function or the new context. As such, we consider the contextual stochastic search paradigm. Here, we want to find good parameter vectors for multiple related tasks, where each task is described by a continuous context vector. Hence, the objective function might change slightly for each parameter vector evaluation. Contextual algorithms have been investigated in the field of policy search. However, contextual policy search algorithms typically suffer from premature convergence and perform unfavourably in comparison with state of the art stochastic search methods. In this paper, we investigate a contextual stochastic search algorithm known as Contextual Relative Entropy Policy Search (CREPS), an information-theoretic algorithm that can learn for multiple tasks simultaneously. We extend that algorithm with a covariance matrix adaptation technique that alleviates the premature convergence problem. We call the new algorithm Contextual Relative Entropy Policy Search with Covariance Matrix Adaptation (CREPS-CMA). We will show that CREPS-CMA outperforms the original CREPS by orders of magnitude. We illustrate the performance of CREPS-CMA on several contextual tasks, including a complex simulated robot kick task.

Abstract
The utility of computer-based teaching-learning systems is generally accepted but several relevant issues remain unsolved in the design of those systems, namely, how to adapt to a learner's specific needs; how to plan corrective feedback; how to fit teaching-learning-assessment techniques to a specific educational context; how to choose the educational tools more appropriate to a teaching-learning-assessment method; how to choose a language to express a pedagogical model; how to adequate the teaching-learning-assessment activities deployment to a specific educational format (distance, face-to-face or blending learning). The aim of this paper is threefold: first, it surveys the most relevant computer-based teaching-learning systems since 1960. Second, it describes the learning design paradigm supported by specific modelling languages. Finally, it presents some reflections on educational material design, more specifically teaching-learning activities, that should be considered by teachers. Those considerations aim at bridging the gap between relevant theoretical aspects and the teachers' daily activities in the design of teaching-learning scenarios.

Abstract
In University Management, one of the recurring problems that each department has to solve, each year, is the Teaching Service Distribution (TSD) or Teaching Assignment Problem (TAP). The problem of TSD consist to assign teachers to courses classes - lectures, tutorials, practical or laboratory - taking into account these preferences and qualifications for teaching. This is a crucial stage, since it is almost imperative that the TSD is fully defined before the process of schedules generating. However, most institutions of higher education, don't have a specific software tool to support the process of TSD. In this paper we propose a new approach for solving the TSD consisting on the formulation of the problem as a distributed scheduling problem with the formation with coalitions formation. The problem is solved in the context of a multi-agent system where the real agents are modeled by computational agents, with their interests, but may cooperate in alliance groups.

Abstract
For the internet of things (IoT) to fully emerge, it is necessary to design a suitable system architecture and specific protocols for this environment. The former to provide horizontal solutions, breaking away the current paradigm of silos solutions, and thus, allowing the creation of open and interoperable systems; while the latter will offer efficient and scalable communications. This paper presents the latest standards and ongoing efforts to develop specific protocols for IoT. Furthermore, this paper presents a new system, with the most recent standards for IoT. Its design, implementation and evaluation will be also described. The proposed system is based on the latest ETSI M2M specification (ETSI TC M2M in ETSI TS 103 093 V2.1.1.http://www.etsi.org/deliver/etsi_ts/103000_103099/103093/02.01.01_60/ts_103093v020101p.pdf, 2013b) and the MQTT protocol (IBM, Eurotech in MQTT V3.1 Protocol Specification pp 1-42, http://public.dhe.ibm.com/software/dw/webservices/ws-mqtt/MQTT_V3.1_Protocol_Specific. pdf, 2010). With this solution it is possible to show how we can create new applications to run over it and the importance of designing specifically tailored for IoT communication protocols in order to support real-time applications.

Abstract
This paper proposes a platform which allows to evaluate the performance of an Embodied Conversational Agent and the influence of several distinct multimedia elements. Using the "Imitation Game" as described by Alan Turing as a starting point and motivation, the platform allows the configuration of several features which influence the Virtual environment. Through a series of user tests and surveys that are performed using the platform, it is possible to gather data that helps measure the efficiency of each feature.