157 resultados para type inference
em Cambridge University Engineering Department Publications Database
Resumo:
Intracellular replication within specialized vacuoles and cell-to-cell spread in the tissue are essential for the virulence of Salmonella enterica. By observing infection dynamics at the single-cell level in vivo, we have discovered that the Salmonella pathogenicity island 2 (SPI-2) type 3 secretory system (T3SS) is dispensable for growth to high intracellular densities. This challenges the concept that intracellular replication absolutely requires proteins delivered by SPI-2 T3SS, which has been derived largely by inference from in vitro cell experiments and from unrefined measurement of net growth in mouse organs. Furthermore, we infer from our data that the SPI-2 T3SS mediates exit from infected cells, with consequent formation of new infection foci resulting in bacterial spread in the tissues. This suggests a new role for SPI-2 in vivo as a mediator of bacterial spread in the body. In addition, we demonstrate that very similar net growth rates of attenuated salmonellae in organs can be derived from very different underlying intracellular growth dynamics.
Resumo:
In technological superconductors, the Lorentz force on the flux vortices is opposed by inhomogeneous pinning and so the critical current may be controlled by a combination of vortex entanglement, cutting, and cross-joining. To understand the roles of these processes we report measurements of structures in which a weak pinning layer is sandwiched between two strongly pinning leads. Quantitative modeling of the results demonstrates that in such systems the critical current is limited by the deformation of individual vortices and not by subsequent cross-joining processes.
Insulin analog preparations and their use in children and adolescents with type 1 diabetes mellitus.
Resumo:
Standard or 'traditional' human insulin preparations such as regular soluble insulin and neutral protamine Hagedorn (NPH) insulin have shortcomings in terms of their pharmacokinetic and pharmacodynamic properties that limit their clinical efficacy. Structurally modified insulin molecules or insulin 'analogs' have been developed with the aim of delivering insulin replacement therapy in a more physiological manner. In the last 10 years, five insulin analog preparations have become commercially available for clinical use in patients with type 1 diabetes mellitus: three 'rapid' or fast-acting analogs (insulin lispro, aspart, and glulisine) and two long-acting analogs (insulin glargine and detemir). This review highlights the specific pharmacokinetic properties of these new insulin analog preparations and focuses on their potential clinical advantages and disadvantages when used in children and adolescents with type 1 diabetes mellitus. The fast-acting analogs specifically facilitate more flexible insulin injection timing with regard to meals and activities, whereas the long-acting analogs have a more predictable profile of action and lack a peak effect. To date, clinical trials in children and adolescents have been few in number, but the evidence available from these and from other studies carried out in adults with type 1 diabetes suggest that they offer significant benefits in terms of reduced frequency of nocturnal hypoglycemia, better postprandial blood glucose control, and improved quality of life when compared with traditional insulins. In addition, insulin detemir therapy is unique in that patients may benefit from reduced risk of excessive weight, particularly during adolescence. Evidence for sustained long-term improvements in glycosylated hemoglobin, on the other hand, is modest. Furthermore, alterations to insulin/insulin-like growth factor I receptor binding characteristics have also raised theoretical concerns that insulin analogs may have an increased mitogenic potential and risk of tumor development, although evidence from both in vitro and in vivo animal studies do not support this assertion. Long-term surveillance has been recommended and further carefully designed prospective studies are needed to evaluate the overall benefits and clinical efficacy of insulin analog therapy in children and adolescents with type 1 diabetes.
Resumo:
The inhomogeneous Poisson process is a point process that has varying intensity across its domain (usually time or space). For nonparametric Bayesian modeling, the Gaussian process is a useful way to place a prior distribution on this intensity. The combination of a Poisson process and GP is known as a Gaussian Cox process, or doubly-stochastic Poisson process. Likelihood-based inference in these models requires an intractable integral over an infinite-dimensional random function. In this paper we present the first approach to Gaussian Cox processes in which it is possible to perform inference without introducing approximations or finitedimensional proxy distributions. We call our method the Sigmoidal Gaussian Cox Process, which uses a generative model for Poisson data to enable tractable inference via Markov chain Monte Carlo. We compare our methods to competing methods on synthetic data and apply it to several real-world data sets. Copyright 2009.
Resumo:
The inhomogeneous Poisson process is a point process that has varying intensity across its domain (usually time or space). For nonparametric Bayesian modeling, the Gaussian process is a useful way to place a prior distribution on this intensity. The combination of a Poisson process and GP is known as a Gaussian Cox process, or doubly-stochastic Poisson process. Likelihood-based inference in these models requires an intractable integral over an infinite-dimensional random function. In this paper we present the first approach to Gaussian Cox processes in which it is possible to perform inference without introducing approximations or finite-dimensional proxy distributions. We call our method the Sigmoidal Gaussian Cox Process, which uses a generative model for Poisson data to enable tractable inference via Markov chain Monte Carlo. We compare our methods to competing methods on synthetic data and apply it to several real-world data sets.
Resumo:
A novel framework is provided for very fast model-based reinforcement learning in continuous state and action spaces. It requires probabilistic models that explicitly characterize their levels of condence. Within the framework, exible, non-parametric models are used to describe the world based on previously collected experience. It demonstrates learning on the cart-pole problem in a setting where very limited prior knowledge about the task has been provided. Learning progressed rapidly, and a good policy found after only a small number of iterations.