The Internet has become one of the most widely used forms of communication available. Many applications used on the Internet require the user to interact constantly with the network. For example web browsing where the user will expect the browser to respond quickly, to finish loading pages quickly and to do all of this at an equal level for all users. The network's performance is dependant on the protocols it uses and how the resources of the network are distributed. This is why TCP (Transmission Control Protocol) is one of the most important protocols, because it controls the amount of data entering the network and provides reliability to most interactive applications. The thesis starts by introducing a basic TCP model which is later extended to model the effects of burstiness produced by TCP. Burstiness can cause a routers buffer to unnecessarily overflow. These overflows cause TCP connections to under-utilise link bandwidth because of unnecessary packet retransmissions. A model to define a quantitative measure of both burstiness and throughput of a system of TCP connections is introduced. The model gives insight into how the TCP protocol causes burstiness and can be used to find scenarios where burstiness is decreased. This helps to improve the utilization of links by reducing the burstiness of protocols. An important performance metric for interactive traffic is user perceived delay, the delay that an end user would encounter when using an application. An example of user perceived delay is the time a user waits before a HTML web page starts loading. The retransmission delays are the most important type of delay for interactive traffic because they are usually very large. A dynamic priority RED Queue (DPRQ) is introduced which changes the priority of the queues based on the goodput (throughput of succesfully transmitted packets) threshold of the interactive traffic. Using dynamic priority allows packet loss to be reduced by up to eight times for interactive traffic, which intern reduces retransmission delay. Fairness measures how equally network resources are allocated amongst different connections. When a link with TCP connections is overloaded each connection on the link will reduce its throughput to allow all the connections to have approximately equal load. This does not take into account that other links may be under utilized. The fairness issue is addressed by introducing Multipath TCP (MATCP) which allows path selection to occur at the TCP layer. This allows each unique flow to take a different path, instead of all the flows of one source using the same path. Using MATCP, a finer grain of load-balancing can be achieved and the complexity and state required in the network is greatly reduced. Two analytic models are provided in chapters three and four, which investigate slow start and TCP burstiness. In chapter five the DPRQ queue is introduced to reduce user perceived delay. An analytic model of the DPRQ is provided and verified through experimental simulation. In chapter six an analytic model of Multipath TCP is provided, which is also verified by simulation.

No description available.

Nous nous intéressons à la fois à la modélisation des fréquences des mots dans les collections textuelles et aux modèles probabilistes de recherche d'information (RI). Concernant les modèles statistiques de fréquences de mots, nous portons notre attention sur l'étude du phénomène de rafale (burstiness). Nous établissons une propriété sur les distributions de probabilité caractérisant leur capacité à modéliser ce phénomène et nous étudions ensuite les distributions Beta Negative Binomial et Log-Logistique pour la modélisation des fréquences de mots. Nous portons ensuite notre attention sur les modèles probabilistes de RI et leur propriétés fondamentales. Nous pouvons montrer que les modèles classiques ne reposent pas sur des lois de probabilité en rafale, même si des propriétés fondamentales comme la concavité des modèles permettent implicitement de le prendre en compte. Nous introduisons ensuite une nouvelle famille de modèles probabiliste pour la recherche d'information, fondé sur la notion d'information de Shannon et qui permet d'établir un lien conséquent entre les propriétés importantes des modèles de RI et le phénomène de rafale. Enfin, nous étudions empiriquement et théoriquement les modèles de rétro-pertinence. Nous proposons un cadre théorique qui permet ainsi d'expliquer leurs caractéristiques empiriques et leur performances. Ceci permet entre autres de mettre en avant les propriétés importantes des modèles de retro-pertinence et de montrer que certains modèles de référence sont déficients. / The present study deals with word frequencies distributions and their relation to probabilistic Information Retrieval (IR) models. We examine the burstiness phenomenon of word frequencies in textual collections. We propose to model this phenomenon as a property of probability distributions and we study the Beta Negative Binomial and Log-Logistic distributions to model word frequencies. We then focus on probabilistic IR models and their fundamental properties. Our analysis reveals that probability distributions underlying most state-of-the-art models do not take this phenomenon into account , even if fundamental properties of IR models such as concavity enable implicitly to take it into account. We then introduce a novel family of probabilistic IR model, based on Shannon information. These new models bridge the gap between significant properties of IR models and the burstiness phenomenon of word frequencies. Lastly, we study empirically and theoretically pseudo relevance feedback models. We propose a theoretical framework which explain well the empirical behaviour and performance of pseudo relevance feedback models. Overall, this highlights interesting properties for pseudo relevance feedback and shows that some state-of-the-art model are inadequate.

Approximately 20% of the body’s energy consumption is ongoingly consumed by the brain, where the main part is due to the neural activity, which is only increased slightly when doing a demanding task. This ongoingly neural activity are studied with the so called resting-state fMRI, which mean that the neural activity in the brain is measured for participants with no specific task. These studies have been useful to understand the neural function and how the neural networks are constructed and cooperate. This have also been helpful in several clinical research, for example have differences been identified between bipolar disorder and major depressive disorder. Recent research has focused on temporal properties of the ongoing activity and it is well known that neural activity occurs in bursts. In this study, resting-state fMRI data and temporal graph theory is used to develop a point based method (PBM) to quantify these bursts at a nodal level. By doing this, the bursty pattern can be further investigated and the nodes showing the most bursty pattern (i.e hubs) can be identified. The method developed shows a robustness regarding several different aspects. In the method is two different variance threshold algorithms suggested. One local variance threshold (LVT) based on the individual variance of the edge time-series and one global variance threshold (GVT) based on the variance of all edges time-series, where the GVT shows the highest robustness. However, the choice of threshold needs to be adapted for the aims of the current study. Finally, this method ends up in a new measure to quantify this bursty pattern named bursty centrality. The derived temporal graph theoretical measure was correlated with traditional static graph properties used in resting state and showed a low but significant correlation. By applying this method on resting-state fMRI data for 32 young adults was it possible to identify regions of the brain that showed the most dynamic properties, these regions differed between the two thresholding algorithms