Panos M. Pardalos, Industrial and Systems Engineering, University of Florida and Director, Centre for Applied Optimization
Title: Data Mining and Optimization Heuristics for Massive Networks
Abstract: In recent years, data mining and optimization heuristics have been used to analyze many large (and massive) data-sets that can be represented as a network. In these networks, certain attributes are associated with vertices and edges. This analysis often provides useful information about the internal structure of the datasets they represent. We are going to discuss our work on several networks from telecommunications (call graph), financial networks (market graph), social networks, and neuroscience.
In addition, we are going to present recent results on critical element selection. In network analysis, the problem of detecting subsets of elements important to the connectivity of a network (i.e., critical elements) has become a fundamental task over the last few years. Identifying the nodes, arcs, paths, clusters, cliques, etc., that are responsible for network cohesion can be crucial for studying many fundamental properties of a network.
Bio: Panos M. Pardalos serves as Distinguished Professor of
Industrial and Systems Engineering at the University of Florida. He is also an affiliated faculty member of the Computer and Information Science
Department, the Hellenic Studies Center, and the Biomedical Engineering Program. He is also the Director of the Center for Applied Optimization. Dr.
Pardalos is a world leading expert in global and combinatorial optimization. His recent research interests include network design problems, optimization in telecommunications, e-commerce, data mining, biomedical applications, and massive computing.
Pablo Moscato, School of Electrical Engineering and Computer Science, University of Newcastle
Title: Personalized Information-based Medicine: Huge challenges, massive opportunities and some lessons learned
Abstract: A recent report from the McKinsey Global Institute highlights the top six disruptive technologies with highest economic impact: mobile internet, automation of knowledge work, Internet of Things, Cloud, Advanced robotics and Autonomous and near-autonomous vehicles. A close seventh is at the core of information-based medicine, next-generation genomics. These seven technologies account for an estimated value which is at least 28 trillion US dollars a year.
All of them share with Information-based Medicine the need of analyzing large datasets, with "Big Data" being the current buzzword. As such, the need of querying a large variety of digital data and the use of artificial-intelligence and optimization software to find novel insights is not considered a separate technology, but a omnipresent requirement across all technologies.
Personalized Medicine aims at putting the best interests of the patient/individual, at the centre of all medical decisions, institutional practices, and/or drugs and treatments that necessarily be "tailored" to the individual profile. Clearly next-generation genomics is pertinent here, but the automation of knowledge work will also prove vital.
These two perspectives for the future of Medicine should contribute to each other. The novel technologies generate an ocean of data, but without strategic approaches for knowledge reuse they do not deliver for the promise. The huge perceived challenges generally involve large optimization. However, the implicit challenge is the development of new mathematical models that contemporize the needs of personalized medicine, who aims at the best diagnostic and treatment, and Information-based Medicine, with the needs of institutions/governments that aim at delivering the best health policies while minimizing global intervention costs operating under budget constraints.
Bio: Australian Research Council Future Fellow Prof. Pablo Moscato is the founding co-director of the Priority Research Centre for Bioinformatics, Biomarker Discovery and Information-based Medicine (2006-) and the funding director of Newcastle Bioinformatics Initiative (2002-2006) of The University of Newcastle. He is also Chief Investigator of the Australian Research Council Centre in Bioinformatics.
Prof. Moscato has been working in Evolutionary Computation for 25 years, and in heuristic methods for Operations Research problems since 1985. His work and ideas have been highly influential in a large number of scientific and technological fields and his manuscripts have been cited more than 4,211 times (data from Google Scholar). The journal "Memetic Computing" is largely dedicated to a methodology he championed since early work with Mike Norman at Caltech in 1988 (memetic algorithms). He is one of Australia's most cited computer scientists.
In the past seven years he has introduced a unifying hallmark of cancer progression based on the changes of information theory quantifiers, developed a novel mathematical model and an associated solution procedure based on combinatorial optimization techniques to identify drug combinations for cancer therapeutics. He has also identified proteomic signatures to predict years in advance the clinical symptoms of Alzheimer's Disease among other `firsts'. His current fellowship supports him for four years (2012-2016) to develop memetic algorithms for multiobjective optimization problems in clinical bioinformatics for personalized and translational medicine.
Steven Kimbrough, Wharton School, University of Pennsylvania
Title: Solution Pluralism, Deliberation, and Metaheuristics
Abstract: We wish to challenge two verities in the MS/OR community as a way of promoting a conceptual shift regarding optimization and metaheuristics. The first verity, roughly, is that given a constrained optimization model the primary problem posed is to find an optimal solution to the model. We call this the goal of optimization modeling verity. The second, roughly, is that exactly optimal solutions (to optimization problems) are always preferred, but heuristically optimal solutions are acceptable if exactly optimal solutions are not available. We call this the justification of heuristics (in optimization) verity. Our challenges to these verities lie not in denying their truth so far as they go, but in denying that they have gone far enough.
The conceptual shift we propose may be described as solution pluralism for deliberation with models. In a nutshell: Given an optimization model, it is possible to define a set of solutions of interest (SoIs), which if well sampled would be valuable for deliberation for decision making (based on the model); further, while the problem of obtaining good samples of the SoIs is challenging, metaheuristics bid fair to be the favored approach and can be shown to be effective in many cases. Among the reasons why SoIs can usefully be defined is that, as is well known, constraints not based in logic often have somewhat arbitrarily chosen coefficient values and these constraints are amenable to adjustment if the increase in profit or decrease in cost is sufficiently large. The talk elaborates upon and illustrates these points.
Bio: Steven Kimbrough is a Professor at The Wharton School,
University of Pennsylvania. His main research interests are in the fields of electronic commerce (and formal languages for business communication),
knowledge and information management, and computational rationality. His active research areas include: computational approaches to belief revision and
nonmonotonic reasoning, formal languages for business communication, evolutionary computation (including genetic algorithms and genetic programming), and information discovery in unstructured and semi-structured data bases (e.g., text). He was principal investigator for the U. S. Coast Guard's KSS
(knowledge-based decision support systems) project, and co-principal investigator on the Logistics DSS project, which is part of DARPA's Advanced
Logistics Program. He was most recently Principal Investigator in the NSF-funded project "Working Memory and Adaptive Choice Behavior."
Michael Trick, Tepper School of Business, Carnegie Mellon University
(LARC-sponsored invited speaker)
Title: Combining Optimization and Metaheuristics in Sports Scheduling
Abstract: Sports scheduling has grown greatly in importance in recent years as more professional and amateur leagues move away from ad-hoc scheduling to optimization-based approaches. Real leagues, however, are generally of a size and structure that precludes straightforward modeling and optimization. Instead, techniques need to be developed that combine optimization with metaheuristic approaches. This has led to a variety of methods for real sports leagues, including large-neighborhood search, genetic algorithms with optimization-based crossovers, and Benders-guided greedy approaches. I will illustrate these methods with examples drawn from professional leagues, such as (U.S.) Major League Baseball, and university leagues. I will also outline the effect that better predictive modeling can have on finding schedules more profitable for the teams and leagues.
Bio: Michael Trick is the Harry B. and James H. Higgins Professor of Operations Research and Senior Associate Dean at the Tepper School of Business, Carnegie Mellon. His research interests are in integer and constraint programming, with a particular emphasis on applications in scheduling and in social choice. In 2002, he was President of INFORMS, the 14,000-member U.S. based operations research society. In his consulting work, he has worked with the United States Postal Service, the U.S. Internal Revenue Service, and many sports leagues, including Major League Baseball, with which he has worked for more than 15 years on scheduling issues. He is the author of more than 50 publications and is the editor or co-editor of six volumes of refereed papers.