The number of R extension packages available from the CRAN repository has tremendously grown over the past 10 years. We look at this phenomenon in more detail, and discuss some of its consequences. In particular, we argue that the statistical computing community needs a more common understanding of software quality, and better domain-specific semantic resources.

Submitted by Luciana Ferreira (lucgeral@gmail.com) on 2018-11-05T14:23:30Z No. of bitstreams: 2 Dissertação - Phelipe Alves de Souza - 2018.pdf: 2593287 bytes, checksum: 75272a4ac609ad844ce539216911cb72 (MD5) license_rdf: 0 bytes, checksum: d41d8cd98f00b204e9800998ecf8427e (MD5) / Approved for entry into archive by Luciana Ferreira (lucgeral@gmail.com) on 2018-11-05T14:30:42Z (GMT) No. of bitstreams: 2 Dissertação - Phelipe Alves de Souza - 2018.pdf: 2593287 bytes, checksum: 75272a4ac609ad844ce539216911cb72 (MD5) license_rdf: 0 bytes, checksum: d41d8cd98f00b204e9800998ecf8427e (MD5) / Made available in DSpace on 2018-11-05T14:30:42Z (GMT). No. of bitstreams: 2 Dissertação - Phelipe Alves de Souza - 2018.pdf: 2593287 bytes, checksum: 75272a4ac609ad844ce539216911cb72 (MD5) license_rdf: 0 bytes, checksum: d41d8cd98f00b204e9800998ecf8427e (MD5) Previous issue date: 2018-10-05 / Coordenação de Aperfeiçoamento de Pessoal de Nível Superior - CAPES / There are great expectations in CRAN and network virtualization (NFV) technologies, and especially in view of the potential they have to accelerate the deployment of new services while lowering the costs of network operators. Several papers discussed the benefits of deploying a new network infrastructure with such technologies, but only a few investigated how the transition from a legacy network could be. In this context, there is a relevant problem that involves three main issues: 1) which network locations should be updated; 2) how to update the selected location, \ie, to fully virtualized or not; and 3) who should attend virtualized sites. These issues are influenced by the level of centralization employed in a given access network (RAN). Here we propose two optimization models and two heuristics that allow the decision maker to define the desired level of centralization and to evaluate its impact on some metrics such as the investment needed and the level of centralization actually achieved. The models show how the investment should be applied according to the level of centralization and the relative cost between the different resources. Our heuristics present similar performance to the exact approach for relatively small scenarios of the problem, but are able to solve topologies of networks with large number of vertices and maintain a satisfactory solution close to the ideal. / Existem grandes expectativas nas tecnologias de centralização (CRAN) e de virtualização de rede (NFV), e especialmente diante do potencial que têm de acelerar a implantação de novos serviços e, ao mesmo tempo, diminuir os custos das operadoras de redes. Vários trabalhos discutiram os benefícios de se implantar uma nova infraestrutura de rede, com tais tecnologias, mas apenas alguns investigaram como poderia ser a transição a partir de uma rede legada. Nesse contexto, existe um problema relevante que envolve três questões principais: 1) quais locais da rede devem ser atualizados; 2) como atualizar o local selecionado, \ie, para totalmente virtualizado ou não; e 3) quem deve atender aos locais virtualizados. Essas questões são influenciadas pelo nível de centralização empregado em uma determinada rede de acesso (RAN). Aqui, propomos dois modelos de otimização e duas heurísticas que permitem ao tomador de decisão definir o nível de centralização desejado e avaliar seu impacto em algumas métricas, tais como o investimento necessário e o nível de centralização efetivamente alcançado. Os modelos mostram como o investimento deve ser aplicado de acordo com o nível de centralização e o custo relativo entre os diferentes recursos. Nossas heurísticas apresentam desempenho semelhante à abordagem exata para cenários relativamente pequenos do problema, mas são capazes de resolver topologias de redes com grande número de vértices e manter uma solução satisfatória próxima ao ideal.

Depuis plusieurs années, les opérateurs de téléphonie mobile sont confrontés à une croissance considérable du trafic de données mobiles. Dans un tel contexte, la technologie Cloud Radio Access Network (CRAN) qui intègre les solutions de Cloud Computing aux réseaux d’accès radio est considérée comme une nouvelle architecture pour les futures générations de réseaux 5G. L’approche CRAN permet une optimisation globale des fonctions de traitement en bande de base du signal et de la gestion des ressources radio pour l’ensemble des RRH et des utilisateurs. Parallèlement, les réseaux hétérogènes (HetNets) ont été proposés pour augmenter efficacement la capacité et la couverture du réseau 5G tout en réduisant la consommation énergétique. En combinant les avantages du Cloud avec ceux des réseaux HetNets, le concept de réseaux H-CRAN (Heterogeneous Cloud Radio Access Networks) est né et est considéré comme l’une des architectures les plus prometteuses pour répondre aux exigences des futurs systèmes. Plus particulièrement, nous abordons le problème important de l’optimisation jointe de l’association utilisateur-RRH et de la solution de beamforming sur la liaison descendante d’un système H-CRAN. Nous formulons un problème de maximisation du débit total du système sous des contraintes de mobilité et d’imperfection de CSI (Channel State Information). Notre principal défi consiste à concevoir une solution capable de maximiser le débit tout en permettant, contrairement aux autres solutions de référence, de réduire la complexité de calcul, et les coûts de signalisation et de feedback CSI dans divers environnements. Notre étude commence par proposer un algorithme Hybride, qui active périodiquement des schémas de clustering dynamiques et statiques pour aboutir à un compromis satisfaisant entre optimalité et le coût en complexité et signalisation CSI et réassociation. L’originalité de l’algorithme Hybride réside aussi dans sa prise en compte de la dimension temporelle du processus d’allocation sur plusieurs trames successives plutôt que son optimalité (ou sous-optimalité) pour la seule trame d’ordonnancement courante. De plus, nous développons une analyse des coûts de l’algorithme en fonction de plusieurs critères afin de mieux appréhender le compromis entre les nombreux paramètres impliqués. La deuxième contribution de la thèse s’intéresse au problème sous la perspective de la mobilité utilisateur. Deux variantes améliorées de l’algorithme Hybride sont proposées : ABUC (Adaptive Beamforming et User Clustering), une version adaptée à la mobilité des utilisateurs et aux variations du canal radio, et MABUC (Mobility-Aware Beamforming et User Clustering), une version améliorée qui règle dynamiquement les paramètres de feedback du CSI (périodicité et type de CSI) en fonction de la vitesse de l’utilisateur. L’algorithme MABUC offre de très bonnes performances en termes de débit cible tout en réduisant efficacement la complexité et les coûts de signalisation CSI. Dans la dernière contribution de la thèse, nous approfondissons l’étude en explorant l’optimisation automatique des paramètres d’ordonnancement du CSI. Pour ce faire, nous exploitons l’outil de l’apprentissage par renforcement afin d’optimiser les paramètres de feedback CSI en fonction du profil de mobilité individuelle des utilisateurs. Plus spécifiquement, nous proposons deux modèles d’apprentissage. Le premier modèle basé sur un algorithme de type Q-learning a permis de démontrer l’efficacité de l’approche dans un scénario à taille réduite. Le second modèle, plus scalable car basé sur une approche Deep Q-learning, a été formulé sous la forme d’un processus de type POMDP (Partially observable Markov decision process). Les résultats montrent l’efficacité des solutions qui permettent de sélectionner les paramètres de feedback les plus adaptés à chaque profil de mobilité, même dans le cas complexe où chaque utilisateur possède un profil de mobilité différent et variable dans le temps. / Recently, mobile operators have been challenged by a tremendous growth in mobile data traffic. In such a context, Cloud Radio Access Network (CRAN) has been considered as a novel architecture for future wireless networks. The radio frequency signals from geographically distributed antennas are collected by Remote Radio Heads (RRHs) and transmitted to the cloud-centralized Baseband Units (BBUs) pool through fronthaul links. This centralized architecture enables a global optimization of joint baseband signal processing and radio resource management functions for all RRHs and users. At the same time, Heterogeneous Networks (HetNets) have emerged as another core feature for 5G network to enhance the capacity/coverage while saving energy consumption. Small cells deployment helps to shorten the wireless links to end-users and thereby improving the link quality in terms of spectrum efficiency (SE) as well as energy efficiency (EE). Therefore, combining both cloud computing and HetNet advantages results in the so-called Heterogeneous-Cloud Radio Access Networks (H-CRAN) which is regarded as one of the most promising network architectures to meet 5G and beyond system requirements. In this context, we address the crucial issue of beamforming and user-to-RRH association (user clustering) in the downlink of H-CRANs. We formulate this problem as a sum-rate maximization problem under the assumption of mobility and CSI (Channel State Information) imperfectness. Our main challenge is to design a framework that can achieve sum-rate maximization while, unlike other traditional reference solutions, being able to alleviate the computational complexity, CSI feedback and reassociation signaling costs under various mobility environments. Such gain helps in reducing the control and feedback overhead and in turn improve the uplink throughput. Our study begins by proposing a simple yet effective algorithm baptized Hybrid algorithm that periodically activates dynamic and static clustering schemes to balance between the optimality of the beamforming and association solutions while being aware of practical system constraints (complexity and signaling overhead). Hybrid algorithm considers time dimension of the allocation and scheduling process rather than its optimality (or suboptimality) for the sole current scheduling frame. Moreover, we provide a cost analysis of the algorithm in terms of several parameters to better comprehend the trade-off among the numerous dimensions involved in the allocation process. The second key contribution of our thesis is to tackle the beamforming and clustering problem from a mobility perspective. Two enhanced variants of the Hybrid algorithm are proposed: ABUC (Adaptve Beamforming and User Clustering), a mobility-aware version that is fit to the distinctive features of channel variations, and MABUC (Mobility-Aware Beamforming and User Clustering), an advanced version of the algorithm that tunes dynamically the feedback scheduling parameters (CSI feedback type and periodicity) in accordance with individual user velocity. MABUC algorithm achieves a targeted sum-rate performance while supporting the complexity and CSI signaling costs to a minimum. In our last contribution, we propose to go further in the optimization of the CSI feedback scheduling parameters. To do so, we take leverage of reinforcement learning (RL) tool to optimize on-the-fly the feedback scheduling parameters according to each user mobility profile. More specifically, we propose two RL models, one based on Q-learning and a second based on Deep Q-learning algorithm formulated as a POMDP (Partially observable Markov decision process). Simulation results show the effectiveness of our proposed framework, as it enables to select the best feedback parameters tailored to each user mobility profile, even in the difficult case where each user has a different mobility profile.

[pt] O emprego conjunto de Redes de Acesso por Rádio em Nuvem (CRANs) e sistemas de múltiplas entradas e múltiplas saídas (MIMO) de larga escala é uma solução chave para atender aos requisitos da quinta geração (5G) de redes sem fio. No entanto, alguns desafios ainda precisam ser superados como a redução do consumo de energia do sistema, a capacidade limitada dos links fronthaul e a redução dos custos de implantação e operação. Embora seja prejudicial para o desempenho do sistema, a quantização em baixa resolução é proposta como uma solução para estes desafios. Portanto, técnicas que melhoram o desempenho de sistemas quantizados grosseiramente são necessárias. Em sistemas móveis, os ADCs geralmente são precedidos por um controle de ganho automático (AGC). O AGC trabalha moldando a amplitude do sinal recebido dentro do intervalo do quantizador para usar eficientemente a resolução. A fim de solucionar esses problemas, esta dissertação apresenta uma otimização conjunta do AGC, que funciona nas cabeças de rádio remotas (RRHs), e um filtro de recepção linear de baixa resolução consciente (LRA) baseado no mínimo erro quadrático médio (MMSE), que funciona na unidade de nuvem (CU), para sistemas quantizados grosseiramente. Desenvolvemos receptores de cancelamento de interferência lineares e sucessivos (SIC) com base na proposta conjunta de AGC e LRA MMSE (AGC-LRA-MMSE). Uma análise da soma das taxas alcançáveis juntamente com um estudo de complexidade computacional também são realizadas. As simulações mostram que o projeto proposto fornece taxas de erro reduzidas e taxas alcançáveis mais altas do que as técnicas existentes. / [en] The joint employment of Cloud Radio Access Networks (C-RANs) and large-scale multiple-input multiple-output (MIMO) systems is a key solution to fulfill the requirements of the fifth generation (5G) of wireless networks. However, some challenges are still open to be overcome such as the high power consumption of large-scale MIMO systems, which employ a large number of analog-to-digital converters (ADCs), the capacity bottleneck of the fronthaul links and the system cost reduction. Although it often affects the system performance, the low-resolution quantization is a possible solution for these problems. Therefore, techniques that improve the performance of coarsely quantized systems are needed. In mobile applications, the ADCs are usually preceded by an automatic gain control (AGC). The AGC works shaping the received signal amplitude within the quantizer range to efficiently use the ADC resolution. Then, the optimization of an AGC is especially important. In order to present possible solutions for these issues, this thesis presents a joint optimization of the AGC, which works in the remote radio heads (RRHs), and a low-resolution aware (LRA) linear receive filter based on the minimum mean square error (MMSE), which works in the cloud unit (CU), for coarsely quantized large-scale MIMO with CRAN systems. We develop linear and successive interference cancellation (SIC) receivers based on the proposed joint AGC and LRA MMSE (AGCLRAMMSE) approach. An analysis of the achievable sum rates along with a computational complexity study is also carried out. Simulations show that the proposed design provides improved error rates and higher achievable rates than existing techniques.

Climate change coupled with land-use change will likely alter habitats and affect state parameters of the animal populations that dwell in them. Affected parameters are anticipated to include site occupancy and abundance, population range, and phenophase cycles (e.g., arrival dates on breeding grounds for migrant bird species). Detecting these changes will require monitoring many sites for many years, a process that is well suited for an automated system. We developed and tested monitoR, an R package that is designed for long-term, multi-taxa automated passive acoustic monitoring programs. monitoR correctly identified presence for black-throated green warbler and ovenbird in 64% and 72% of the 52 surveys using binary point matching, respectively, and 73% and 72% of the 52 surveys using spectrogram cross-correlation, respectively. Of individual black-throated green warbler song events, 73% of 166 black-throated green warbler songs and 69% of 502 ovenbird songs were identified by binary point matching. Spectrogram cross correlation identified 64% of 166 black-throated green warbler songs and 64% of 502 ovenbird songs. False positive rates were We describe a method to identify the probability of survey presence in a template-based automated detection system using known false positive rates for each template. True and false positive detection rates were observed in 146 training surveys. These probabilities were used in a Bayesian approach that discriminates between detections in occupied surveys and unoccupied surveys. We evaluated this approach in 146 test surveys. A total of 1142 Black-throated green warbler (Setophaga virens) songs were observed in the training surveys and test surveys, which we attempted to locate with 3 different binary point matching templates. When only posterior probabilities greater than 0.5 were considered detections, the average ratio of accurate identifications of survey presence to false positive identifications in 500 bootstrapped samples improved from 1.2:1 using a standard score cutoff approach to 2.8:1 using all 3 templates and a likelihood-based discriminator. With the selected score cutoffs the average true positive and false positive rates for the combined three templates were 0.18 and 0.002, respectively. Automated detection methods are increasingly being used for identification and monitoring of landscape-scale responses to climate change and land-use change. Skepticism of automated acoustic monitoring software is largely due to higher false positive and negative error rates than those in traditional human surveys, but the false positive multiple method occupancy model is capable of estimating detection parameters and occupancy state when one method has occasional false positive detections. We test the accuracy of the model when automated detection of black-throated green warbler is mixed with human detection in 4 recorded surveys at 60 sites. Precision and accuracy are evaluated by simulation, and we use the results to optimize future sampling. In simulation, parameter estimates by the multiple method occupancy model are close to those we computed manually when two surveys are manually analyzed. Our results support the use of the multiple method false positive occupancy model to track detection rates in automated monitoring programs.

One of the cornerstones of the R system for statistical computing is the multitude of contributed packages making an extremely broad range of statistical techniques and other quantitative methods freely available. This study investigates which factors are the crucial determinants responsible for the participation of the package authors in the R project. For this purpose a survey was conducted among R package authors, collecting data on different types of participation in the R project, three psychometric scales (hybrid forms of motivation, work design characteristics, and values), as well as various specie-demographic factors. These data are analyzed using item response theory and generalized linear models, showing that the most important determinants for participation are a hybrid form of motivation and the knowledge characteristics of the work design. Other factors are found to have less impact or influence only specific aspects of participation. (authors' abstract) / Series: Research Report Series / Department of Statistics and Mathematics