Abstract

Abstract
In this paper, we describe a new corpus-named DIRHA-L2F RealCorpus-composed of typical home automation speech interactions in European Portuguese that has been recorded by the INESC-ID's Spoken Language Systems Laboratory (L2F) to support the activities of the Distant-speech Interaction for Robust Home Applications (DIRHA) EU-funded project. The corpus is a multi-microphone and multi-room database of real continuous audio sequences containing read phonetically rich sentences, read and spontaneous keyword activation sentences, and read and spontaneous home automation commands. The background noise conditions are controlled and randomly recreated with noises typically found in home environments. Experimental validation on this corpus is reported in comparison with the results obtained on a simulated corpus using a fully automated speech processing pipeline for two fundamental automatic speech recognition tasks of typical "always-listening" home-automation scenarios: system activation and voice command recognition. Attending to results on both corpora, the presence of overlapping voice-like noise is shown as the main problem: simulated sequences contain concurrent speakers that result in general in a more challenging corpus, while real sequences performance drops drastically when TV or radio is on.

Abstract
We merge the two-sided markets duopoly model of Armstrong (2006) with the nested vertical and horizontal differentiation model of Gabszewicz and Wauthy (2012), which consists of a linear city with different consumer densities on the left and on the right side of the city. In equilibrium, the high-quality platform sells at a higher price and captures a greater market share than the low-quality platform, despite the indifferent consumer being closer to the high-quality platform. The difference between market shares is lower than socially optimal. A perturbation that introduces a negligible difference between the consumer density on the left and on the right side of the city may disrupt existence of equilibrium in the model of Armstrong (2006).

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
Background: Hemiballism may arise as a rare consequence of focal basal ganglia lesions. Pathophysiologically, there is a controversy between the role of the STN as the exclusive lesion localization as opposed to several brain regions in which lesions may induce hemiballism. This is most likely due to a motor circuit affection. Objectives: To study the affection of neural networks in the pathogenesis of hemiballism. Methods: We analysed focal vascular lesions inducing hemiballism (n = 8), their localizations and connectivity profiles. Probabilistic tractography (FSL: http://fsl.fmrib.ox.ac.uk/fsl/) was used to study connectivity. Results: Lesions inducing hemiballism were distributed across several anatomic regions (basal ganglia, thalamus, caudate, internal capsule) without a clear predilection. However, we detected increased connectivity for these lesions toward the STN and mesial cortical motor regions (pre-SMA/SMA). These regions are interconnected via subthalamo-pallido-thalamo-cortical networks. Conclusions: We provide evidence for the involvement of the subthalamo-pallido-thalamic pathways in the pathogenesis of hemiballism, which is consistent with data on experimental hemiballism in animals. Electrophysiological basal ganglia recordings and functional MRI would complement our findings to assess the activation patters within these circuits.

Abstract
Attention-to-task is one of the most important Human cognitive abilities, allowing an individual to selectively focus on a specific issue (among many possible sources) and effectively carry out a task. Without this ability to focus, the individual would constantly switch between stimuli, hardly concluding any task. While attention can be influenced by many internal and external factors, the purpose of this paper is not to analyse them but rather to propose an approach to monitor the attentional behaviour of computer users. The proposed approach may improve the individual's self-awareness as well as the team manager's knowledge about the state of the workforce. It may thus improve the definition of better attention-management strategies, with expected improvements in variables such as on-task behaviour, productivity or work quality.