We would like to invite you to our next interesting Guest Talk by Dr. Violet R. Syrotiuk, Associate Professor at Arizona State University, on Combinatorial Requirements for Large Scale Screening Experiments.
Abstract
The discipline of design and analysis of experiments provides a theoretical basis for experimentation with a long tradition of connections to algebra, geometry, and combinatorics. While experimentation is used for many purposes, our interest is in screening experiments, classically used to identify the factors that significantly impact the system responses.
Certain assumptions and limitations underlie many screening designs. Most designs only consider a set of factors that is “not too large,” with two levels for each factor, and are not able to treat categorical factors. Further, the designs often miss unexpected behaviour resulting from interactions of factors. In the analysis of the experiment, it is often assumed that the corresponding design is balanced, that the system reports in every run, that the direction of a response is known for specific factors, and that the data are normally distributed.
Overcoming such limiting assumptions leads to challenging and novel problems in combinatorics,
which we discuss. We also present results for a large scale screening experiment in a wireless network
testbed and other real-world scenarios. Other applications of this work arise in software testing.
About the Speaker
Dr. Violet R. Syrotiuk
She is Associate Professor at Arizona State University and her work focuses on various topics, including cross-layer protocol design, statistical analysis, and anomaly detection, as well as the use of combinatorics in networking. She has a particular interest in adaptive network protocols and midscale experimental research infrastructure. Dr. Syrotiuk is also recognized as a Senior Scientist in Sustainability.
Photo credit: © ASU