Meeting Agenda - 2000.06.21
Meeting Agenda - 2000.06.21
Mauro Steigleder 
Gauge: A Frame Map Editor
The specification of a frame is an important step in the modeling of anisotropic shading and the use of frame maps can make this process much easier. Unfortunately, no there is no well-known frame map editor. The purpose of this talk is to present a frame map editor currently being developed that can make the process of definition of frame maps more user friendly. Some aspects of visualization and rendering using frame maps are covered.
We will give two twenty-minute practice talks for presentations at the 7th Scandinavian Workshop on Algorithm Theory (SWAT 2000). The first talk introduces the paper "Embeddings of k-connected graphs of pathwidth k" (joint work with Arvind Gupta and Andrzej Proskurowski). It concerns subgraph isomorphism and its generalizations (topological embedding and minor containment) on graphs which have a path-like structure and a lower bound on connectivity. Our contributions include structural properties and resultant polynomial-time algorithms (some with k appearing in the exponent, some with exponent independent of k). The problems become NP-complete under modest relaxation of the constraints of the input graph. These results help to define sharper boundaries between tractability and intractability, both general and fixed-parameter, of the embedding problems.
The second talk introduces the paper "On graph powers for leaf-labeled trees" (joint work with Dimitrios Thilikos). Given information about the leaves of an unknown tree, is it possible to reconstruct the tree? We give a partial answer to this question in the case where the information is a k-leaf power, that is, a graph whose nodes are the leaves of the tree, with an edge between two nodes if and only if there is a path of length k between the corresponding leaves in the tree. For k=3 and k=4, we give O(n^3) algorithms that, given a graph, determine whether or it is a k-leaf power of a tree, and if so, construct one such tree. If we generalize from distances between leaves in a tree to distances between nodes in an arbitrary graph, the problem becomes NP-hard.
Poisson mixed models are useful for accommodating the over-dispersion and correlation often observed among count data. These models are generated from the well-known independent Poisson model by adding normally distributed random effects to the linear predictor, and they are known as Poisson-log-normal mixed models. Unfortunately, a full likelihood analysis for such Poisson mixed models is hampered by the need for numerical integrations. To overcome such integration problems, recently, Jiang (1998, JASA) has suggested a computationally feasible method of simulated moments (MSM) for inference in a wider `generalized linear mixed models' (GLMMs) set-up.
In this talk, I demonstrate in connection with the Poisson mixed model, that the MSM approach, in contrary to the existing claims, does not actually produce consistent estimates for the variance component of the model. I will further discuss a joint quasi-likelihood (JQL) approach which is computationally feasible and at the same time leads to substantial improvement over the MSM approach in terms of relative bias and/or mean squared errors of the estimates. The results of a simulation study on the relative performance of the JQL and the MSM approaches for both large and small samples, will also be discussed.
To be announced.
Given the substantial increases in both trade volume and trade complexity, significant improvements in processing power, scalability and reliability are required to stay competitive in derivatives trading today. I describe a new approach that we are undertaking at MSDW to satisfy these goals based on set of loosely coupled components. I will talk about some of the issues that we face such as scheduling, failover, priorization hetergenous and unreliable hardware and very heterogenous request sizes (which have a range of about 5 orders of magnitude).