Catherine Loader

Smoothing: Local Regression Techniques

Smoothing methods attempt to find functional relationships between different measurements. As in the standard regression setting, the data is assumed to consist of measurements of a response variable, and one or more predictor variables. Standard regression techniques (Chap. III8.) specify a functional form (such as a straight line) to describe the relation between the predictor and response variables. Smoothing methods take a more flexible approach, allowing the data points themselves to determine the form of the fitted curve.

Robust multiclass signaling overload control

We propose multi-class signaling overload control algorithms, for telecommunication switches, that are robust against different input traffic patterns and system upgrades. In order to appropriately measure the system load when several classes of signaling traffic are present, we first introduce the concept of equivalent system load measure that converts the multiple system measures associated with different classes of traffic into a single measure with respect to a pre-defined base class. We use this measure to develop three multi-class overload detection and measurement algorithms. Next, we develop a new algorithm for partitioning the allowable equivalent system load across multiple traffic classes, using a strict priority scheme. Using simulations of call flows from mobile telecommunications standards, we compare different multi-class overload algorithms under a variety of overload conditions. Our simulation results indicate that our algorithm that measures system load using a combination of request acceptance rate and processor occupancy provides highly reactive and robust overload control. Last, for the purpose of making the overload control algorithms more robust, we propose a measurement-based simple regression technique to dynamically estimate key system parameters. We find that estimates derived in this manner converge rapidly to their true values.

New technique for finding needles in haystacks : Geometric approach to distinguishing between a new source and random fluctuations

We propose a new test statistic based on a score process for determining the statistical significance of a putative signal that may be a small perturbation to a noisy experimental background. We derive the reference distribution for this score test statistic; it has an elegant geometrical interpretation as well as broad applicability. We illustrate the technique in the context of a model problem from high-energy particle physics. Monte Carlo experimental results confirm that the score test results in a significantly improved rate of signal detection.

Iteratively Reweighted Generalized Least Squares for Estimation and Testing With Correlated Data: An Inference Function Framework

The focus of this article is on fitting regression models and testing of general linear hypotheses for correlated data using quasi-likelihood based techniques. The class of generalized method of moments or GMMs provides an elegant approach for estimating a vector of regression parameters from a set of score functions. Extending the principle of the GMMs, in the generalized estimating equation framework, leads to a quadratic inference function or QIF approach for the analysis of correlated data. We derive an iteratively reweighted generalized least squares or IRGLS algorithm for finding the QIF estimator and establish its convergence properties. A software library implementing the techniques is demonstrated through several datasets.

Robust router congestion control using acceptance and departure rate measures

In packet switched networks, routers need to implement controls to overcome the problems of reduced performance in the event of congestion. In this work, with a view towards robust implementation, we examine a new active queue management scheme, Acceptance and Departure Rate (ADR) that uses a combination of acceptance rate and departure rate measures to control link congestion. Unlike many existing approaches, ADR does not implicitly or explocity use queue length measures which are not robust to changes in system capacity. The main advantage of our approach being its robustness with varying network parameters including link capacity and network load. We also compare the performance of our approach with other known active queue management schemes.