Les techniques adaptatives de transmission vidéo s’appuient sur un contenu qui est encodé à différents niveaux de qualité et divisé en segments temporels. Avant de télécharger un segment, le client exécute un algorithme d’adaptation pour décider le meilleur niveau de qualité à considérer. Selon les services, ce niveau de qualité doit correspondre aux ressources réseaux disponibles, mais aussi à d’autres éléments comme le mouvement de tête d’un utilisateur regardant une vidéo immersive (à 360°) afin de maximiser la qualité de la portion de la vidéo qui est regardée. L’efficacité de l’algorithme d’adaptation a un impact direct sur la qualité de l’expérience finale. En cas de mauvaise sélection de segment, un client HTTP/1 doit attendre le téléchargement du prochain segment afin de choisir une qualité appropriée. Dans cette thèse, nous proposons d’utiliser le protocole HTTP/2 pour remédier à ce problème. Tout d’abord, nous nous focalisons sur le service de vidéo en direct. Nous concevons une stratégie de rejet d’images vidéo quand la bande passante est très variable afin d’éviter les arrêts fréquents de la lecture vidéo et l’accumulation des retards. Le client doit demander chaque image vidéo dans un flux HTTP/2 dédié pour contrôler la livraison des images par appel aux fonctionnalités HTTP/2 au niveau des flux concernées. Ensuite, nous optimisons la livraison des vidéos immersives en bénéficiant de l’amélioration de la prédiction des mouvements de têtes de l’utilisateur grâce aux fonctionnalités d’initialisation et de priorité de HTTP/2. Les résultats montrent que HTTP/2 permet d’optimiser l’utilisation des ressources réseaux et de s’adapter aux latences exigées par chaque service. / Adaptive video streaming techniques enable the delivery of content that is encoded at various levels of quality and split into temporal segments. Before downloading a segment, the client runs an adaptation algorithm to determine the level of quality that best matches the network resources. For immersive video streaming this adaptation mechanism should also consider the head movement of a user watching the 360° video to maximize the quality of the viewed portion. However, this adaptation may suffer from errors, which impact the end user’s quality of experience. In this case, an HTTP/1 client must wait for the download of the next segment to choose a suitable quality. In this thesis, we propose to use the HTTP/2 protocol instead to address this problem. First, we focus live streaming video. We design a strategy to discard video frames when the band width is very variable in order so as to avoid the rebuffering events and the accumulation of delays. The customer requests each video frame in an HTTP/2 stream which allows to control the delivery of frames by leveraging the HTTP/2 features at the level of the dedicated stream. Besides, we use the priority and reset stream features of HTTP/2 to optimize the delivery of immersive videos. We propose a strategy to benefit from the improvement of the user’s head movements prediction overtime. The results show that HTTP/2 allows to optimize the use of network resources and to adapt to the latencies required by each service.

A presente tese investigou os efeitos de estresse na latência e acurácia de respostas motoras a estímulos visuais, e a eficiência de treino na reversão de tais efeitos. O embasamento teórico adotado foi o Attentional Control Theory, que prediz que em situações de estresse, o sistema atencional goal-directed fica prejudicado e sobrepujado pelo sistema stimulus-driven. O pênalti no futebol foi adotado como paradigma experimental. Os resultados revelaram que quando estressados participantes ou precisaram de mais tempo para responder ao estímulo visual ou eram incompetente em inibir respostas automáticas incompatíveis com a tarefa. Tais efeitos foram em parte revertidos com sucesso, revelando grandes diferenças individuais que parecem estar relacionadas com controle atencional. Os resultados também revelaram que a simples presença de um estímulo irrelevante para tarefa prejudicou a resposta motora. Os achados ofereceram uma melhor compreensão de como estresse e estímulos irrelevantes afetam o controle motor e de como treinamento pode reverter tais efeitos. / The current thesis investigated the effects of high-pressure on latency and accuracy of motor responses to visual stimuli and the efficiency of task-specific practice in managing these effects. The theoretical framework adopted was Attentional Control Theory, which argues that high-pressure increases stress and thus impairs the goal-directed attentional system. The penalty kick in football was adopted as paradigm. The experiments revealed that when stressed, participants either required more time to respond to the visual stimulus movement or were unable to inhibit automatic responses, resulting in consistent errors. Such effects were in part successfully reverted, revealing individual differences that seem to relate to the individuals tendencies for attentional control. Furthermore, the results showed that the mere presence of a threatening non-target object impaired shooting accuracy. The findings provide a better understanding of how high-pressure and non-target objects affect motor performance and how task-specific practice may revert such effects.

HIV remains an incurable viral infection and a significant global health concern. Despite the advent of antiretroviral therapy, there are 36.9 million recorded cases of HIV worldwide, with an additional 1.8 million new infections recorded in 2017 alone. An HIV cure is therefore one of several priorities within the field, and will require HIV “reservoir” cells—comprised of latently-HIV infected CD4+ T cells and productively-infected, tissue resident macrophages—to be selectively killed in vivo. HIV reservoir cells are rarely found within the peripheral circulation, residing instead within inaccessible tissue sanctuaries. Consequently, their characterization has been limited to in vitro laboratory models. To complicate matters further, a definitive cellular surface marker of HIV infected cells has yet to be identified. Impairment of the type I interferon (IFN1) response has been observed during HIV infection, however, making it a unique intracellular maker of HIV-infected cells. The recent development of oncolytic viruses (OV) designed to selectively kill IFN-defective cancer cells also suggests that these IFN1 defects possess therapeutic value. It was therefore hypothesized that the impairment of the IFN1 response in HIV-infected CD4+ cells and macrophages could serve as a target for oncolytic virus-mediated killing. The induction of several antiviral IFN-stimulated proteins, including PKR and ISG15, was inhibited in HIV-infected monocyte-derived macrophages (MDM) following stimulation with IFNα or a synthetic RNA. Consequently, HIV-infected MDM were more susceptible to infection and killing by the oncolytic Maraba virus, MG1. Importantly, MG1-mediated killing required the presence of replication-competent OV, and could not be potentiated by UV-inactivated MG1 or supernatants from MG1-infected cells. The ability of MG1 to target the HIV reservoir was further confirmed using alveolar macrophages collected from the lungs of cART-suppressed individuals living with HIV. These findings indicate that IFN1 defects are a feature of HIV infected cells, which can be exploited for selective killing by OV. This project is therefore unique in that it demonstrates that HIV reservoir cells can be eradicated in a targeted manner by exploiting an intracellular marker of HIV infection. As MG1-based cancer therapies are currently being explored in Phase I/II clinical trials, there is potential for this approach to be adapted for use within the HIV cure field.

Myostatin is a secreted growth factor of the transforming growth-factor $\beta$ (TGF$\beta$) superfamily, and a powerful negative regulator of muscle mass in vertebrates. As such, there is considerable interest in developing pharmacological agents which inhibit myostatin signalling in order to stimulate muscle growth in the context of pathological muscle wasting. Like other TGF$\beta$ family proteins, myostatin is biosynthesised as an inactive (latent) precursor protein which requires proteolytic processing to liberate the mature bioactive growth factor. To examine the molecular basis of pro-myostatin latency and the mechanism by which it is activated in the extracellular space, I have determined the crystal structure of unprocessed human pro-myostatin and studied the properties of the protein at various stages of activation. Crystallographic analysis of pro-myostatin reveals a unique domain-swapped dimeric structure, with an open V-shaped conformation distinct from the prototypical family member, TGF$\beta$1. Following cleavage of the prodomains by furin, pro-myostatin persists as a stable non-covalent complex which is resistant to the natural inhibitor follistatin and exhibits significantly weaker bioactivity than the mature growth factor. A number of distinct structural features combine to stabilise the interaction between pro and mature domains and in doing so confer latency to the pro-complex. This facilitates a controlled, step-wise process of activation in the extracellular space and contributes to a complex network of regulatory control. The results presented here provide a structural basis for understanding the effect of natural polymorphisms on myostatin function and a starting point for structure-guided development of next generation myostatin inhibitors. As a proof-of-concept, I present preliminary data on prodomain derived stapled peptides as inhibitors of myostatin signalling.

No description available.

A broad understanding of motor control has been achieved through research performed on upper extremity reaching, walking on level ground, and static balance. Though invaluable insights have been achieved under these testing paradigms, inherent limitations result in less being known regarding functional movement in weight-bearing. Gait studies require large numbers of consecutive steps to achieve high reliability, static balance is limited to the goal of no movement, and upper extremity reaching lacks insights into feedback from the vestibular system. Here we describe (and provide a supplemental video of) a system for testing and training the performance of a weight-bearing, visuomotor task in the form of a mini-squat according to a sinusoidal trace on a screen. In this work, we determined that by altering both task movement rate and resistance at the knee, a hierarchy of difficulty was achieved at all ages. As age increases, there is a velocity-error tradeoff; speed of movement is attempted to be maintained while error is sacrifieced. When introducing an unexpected force perturbation (rapid release of the resistance of the squat for less than a second), older adults who are least able to match the frequency of the task experience the greatest error and velocity rates during the perturbation. This exposes a possible deficit in the feedback control system of even healthy older adults, where future studies may determine if early intervention to prevent such changes may prevent future injury and disability. When older and younger adults learned to perform the visuomotor task while performing a simultaneous cognitive task, learning was slowed as complexity of the cognitive task increased. In older adults, a difficult cognitive task inhibited acquisition of the squatting task with no apparent improvement in trial error nor coherence. Upon retesting of the motor task, there was no difference between dual-task and single-task trained ability to consolidate the motor task in both age groups, though those that trained under a dual-task condition were more able to automate the motor task as measured by a smaller dual-task cost. This may indicate that dual-task training leads to freeing cognitive resources from attending to a functional movement so that they may attend to other tasks such as what may be happening in the environment. Finally, executive function as measured by the Flanker Test, explained 80% of the variability of final day visuomotor error, being a possible prognostic factor for dual-task interventions. Future directions will determine if increased automaticity of a mini-squat will lead improvement to overall improved functional mobility and reduced lower extremity injuries when functioning in a busy community.

The importance of energy efficiency has grown alongside awareness of climate change due to the rapid increase of greenhouse gases. With the increasing trend regarding mobile subscribers, it is necessary to prevent an expansion of energy consumption via mobile networks. In this thesis, the energy optimization of the new radio access technology called 5G NR utilizing different sleep states to put base stations to sleep when they are not transmitting data is discussed. Energy savings and file latency with heterogeneous and super dense urban scenarios was evaluated through simulations with different network deployments. An updated power model has been proposed and the sensitivity of the new power model was analyzed by adjusting wake-up time and sleep factors. This showed that careful implementation is necessary when adjusting these parameter settings, although in most cases it did not change the end results by much. Since 5G NR has more potential in energy optimization compared to the previous generation mobile network 4G LTE, up to 4 sleep states was implemented on the NR base stations and one idle mode on LTE base stations. To mitigate unnecessary sleep, deactivation timers are used which decides when to put base stations to sleep. Without deactivation timers, the delay could increase significantly, while with deactivation timers the delay increase would only be a few percent. Up to 42.5% energy could be saved with LTE-NR non-standalone deployment and 72.7% energy with NR standalone deployment compared to LTE standalone deployment, while minimally impacting the delay on file by 1%.

<div>Remote Direct Memory Access, or RDMA, is a technology that allows one computer server to direct access the memory of another server without involving its CPU. Compared with traditional network technologies, RDMA offers several benefits including low latency, high throughput, and low CPU utilization. These features are especially attractive to datacenters, and because of this, datacenters have started to adopt RDMA in production scale in recent years.</div><div>However, RDMA was designed for confined, single-tenant, High-Performance-Computing (HPC) environments. Many of its design choices do not fit datacenters well, and it cannot be readily used by datacenter applications. To use RDMA, current datacenter applications have to build customized software stacks and fine-tune their performance. In addition, RDMA offers limited scalability and does not have good support for resource sharing or protection across different applications.</div><div>This dissertation sets out to seek solutions that can solve issues of RDMA in a systematic way and makes it more suitable for a wide range of datacenter applications.</div><div>Our first task is to make RDMA more scalable, easier to use, and have better support for safe resource sharing in datacenters. For this purpose, we propose to add an indirection layer on top of native RDMA to virtualize its low-level abstraction into a high-level one. This indirection layer safely manages RDMA resources for different datacenter applications and also provide a means for better scalability.</div><div>After making RDMA more suitable for datacenter environments, our next task is to build applications that can exploit all the benefits from (our improved) RDMA. We designed a set of systems that store data in remote persistent memory and let client machines access these data through pure one-sided RDMA communication. These systems lower monetary and energy cost compared to traditional datacenter data stores (because no processor is needed at remote persistent memory), while achieving good performance and reliability.</div><div>Our final task focuses on a completely different and so far largely overlooked one — security implications of RDMA. We discovered several key vulnerabilities in the one-sided communication pattern and in RDMA hardware. We exploited one of them to create a novel set of remote side-channel attacks, which we are able to launch on a widely used RDMA system with real RDMA hardware.</div><div>This dissertation is one of the initial efforts in making RDMA more suitable for datacenter environments from scalability, usability, cost, and security aspects. We hope that the systems we built as well as the lessons we learned can be helpful to future networking and systems researchers and practitioners.</div>

The surge of mobile devices, such as smartphones, and tables, demands additional capacity. On the other hand, Internet-of-Things (IoT) and smart grid, which connects numerous sensors, devices, and machines require ubiquitous connectivity and data security. Additionally, some use cases, such as automated manufacturing process, automated transportation, and smart grid, require latency as low as 1 ms, and reliability as high as 99.99\%. To enhance throughput and support massive connectivity, sharing of the unlicensed spectrum (3.5 GHz, 5GHz, and mmWave) is a potential solution. On the other hand, to address the latency, drastic changes in the network architecture is required. The fifth generation (5G) cellular networks will embrace the spectrum sharing and network architecture modifications to address the throughput enhancement, massive connectivity, and low latency. To utilize the unlicensed spectrum, we propose a fixed duty cycle based coexistence of LTE and WiFi, in which the duty cycle of LTE transmission can be adjusted based on the amount of data. In the second approach, a multi-arm bandit learning based coexistence of LTE and WiFi has been developed. The duty cycle of transmission and downlink power are adapted through the exploration and exploitation. This approach improves the aggregated capacity by 33\%, along with cell edge and energy efficiency enhancement. We also investigate the performance of LTE and ZigBee coexistence using smart grid as a scenario. In case of low latency, we summarize the existing works into three domains in the context of 5G networks: core, radio and caching networks. Along with this, fundamental constraints for achieving low latency are identified followed by a general overview of exemplary 5G networks. Besides that, a loop-free, low latency and local-decision based routing protocol is derived in the context of smart grid. This approach ensures low latency and reliable data communication for stationary devices. To address data security in wireless communication, we introduce a geo-location based data encryption, along with node authentication by k-nearest neighbor algorithm. In the second approach, node authentication by the support vector machine, along with public-private key management, is proposed. Both approaches ensure data security without increasing the packet overhead compared to the existing approaches.

Les modèles de décision conventionnels, basés sur l’utilisation du système saccadique comme modèle sensorimoteur, considèrent typiquement les temps de réaction comme un sous-produit des processus décisionnels, reflétant le temps nécessaire pour prendre une décision. Cependant, des recherches ont montré que les latences saccadiques sont deux fois plus longues que le temps de décision, ainsi que l’organisation de l’environnement affecte les latences saccadiques. Cette thèse propose une interprétation alternative des temps de réaction saccadiques (SRTs) en montrant que les distributions de latences saccadiques peuvent être altérées par leurs propres conséquences. Nous défendons l’hypothèse que les latences saccadiques dépendent de relations fonctionnelles avec leur environnement.Cette thèse a réalisé une analyse fonctionnelle des latences saccadiques. La première étude a évalué s’il était possible de choisir ses propres latences en fonction des contingences de renforcement en cours, dans le but d’explorer l’étendue du contrôle temporel des saccades. L’allocation des latences courtes et longues correspondait au renforcement relativement obtenu, démontrant un contrôle fin des SRTs. La seconde étude a évalué de manière plus approfondie l’effet de conséquences bénéfiques sur les SRTs, en utilisant le phénomène de taille-latence. La procédure de renforcement a été efficace pour manipuler le bénéfice de SRTs plus courts et pour réduire le phénomène de taille-latence. La troisième étude a démontré comment les stimuli antécédents en viennent à contrôler des temps de réaction spécifiques à l’aide d’un apprentissage opérant. Les contingences de renforcement ont induit un contrôle discriminatif des latences entre des stimuli différents. Enfin, la dernière expérience a exploré l’implication des processus d’apprentissage classique dans le contrôle par le stimulus des latences saccadiques. Cette étude pilote met en évidence l’influence de l’environnement et de l’historique d’apprentissage dans le contrôle temporel des saccades.Nos résultats soulignent l’incroyable plasticité du système saccadique, et l’étend au contrôle temporel des saccades. Cette thèse montre qu’un processus général d’apprentissage, basé sur les conséquences fonctionnelles des saccades, peut expliquer de manière parcimonieuse les changements dans les latences saccadiques. Démontrant que la latence est une dimension opérante des saccades, l’organisation de l’environnement contrôle l’organisation temporelle des saccades. / Conventional decision models, based on the saccadic system as a sensorimotor model, typically view reaction time as a byproduct of decisional processes, reflecting the time needed to make a decision. However, research has shown that saccadic latencies are twice as long as the decision time and that the organization of the environment affects saccade latencies. This thesis dissertation provides an alternative view of saccadic reaction times (SRTs) by showing that saccade latency distribution can be altered by their own consequences. We defend that saccadic latency depends on functional relations with its environment.This thesis conducted a functional analysis of saccadic latencies. The first study probed whether it is possible to choose one’s latencies depending on the reinforcement contingencies in force, in order to assess the extent of temporal control with saccades. The allocation of short and long latencies matched the relative reinforcement obtained, demonstrating a fine control of SRTs. The second study further investigated the effect of beneficial consequences on SRTs, using the size-latency phenomenon. The reinforcement procedure was effective in manipulating the benefit of shorter SRTs and reducing the size- latency phenomenon. The third experiment demonstrated how antecedent stimuli come to control specific reaction times through operant learning. Reinforcement contingencies induced discriminative control of latencies between different stimuli. Finally, the last experiment explored the involvement of classical learning processes in stimulus control of saccade latencies. This pilot study highlighted the influence of the environment and learning history in the temporal control of saccades.Our results emphasize the exquisite plasticity of the saccadic system, and extend it to the temporal control of saccades. This thesis shows that a general learning process, based on the functional consequences of saccades, can parsimoniously explain changes in saccadic latency. Demonstrating that latency is an operant dimension of saccades, the organization of the environment controls the temporal organization of saccades.