News Category: ‘DISSERTATION DEFENSES’

Notice and Invitation
Oral Defense of Doctoral Dissertation
Doctor of Philosophy in Computational Sciences and Informatics
Department of Computational and Data Sciences
College of Science
George Mason University

Muhammad N. Baqui
Bachelor of Science, Bangladesh University of Engineering and Technology, 2006
Master of Science, North Dakota State University, 2010

Automated Monitoring of High Density Crowd Events

Tuesday, January 16, 2018, 11:00 a.m.
Research Hall, Room 161

All are invited to attend.

Committee
Rainald Löhner, Dissertation Director
Juan Cebral
Chi Yang
Jan Allbeck

Pedestrian traffic is an important subject of surveillance to ensure public safety and traffic management, which may benefit from intelligent and continuous analysis of Pedestrian traffic videos. State-of- the-art methods for intelligent pedestrian traffic surveillance have a number of limitations in automating and computing useful pedestrian traffic information from closed circuit television (CCTV) images. This work aims to automate and augment the traditional pedestrian traffic surveillance system by introducing four components in a novel integrated framework. A fast and efficient particle image velocimetry (PIV) technique is proposed to yield pedestrian velocities for timely management of pedestrian traffic. A machine learning-based regression model, boosted Ferns, is used to improve pedestrian count estimation: an essential metric for high-density pedestrian traffic analysis. A camera perspective model is proposed to improve the speed and position estimate of pedestrian traffic incorporating camera’s intrinsic and extrinsic parameters. All these functional improvements are integrated in a seamless framework to predict future pedestrian traffic distribution, which is a crucial piece   of information for pedestrian traffic management. The proposed framework is computationally efficient, suitable for multiple camera feeds with high-density pedestrian traffic videos, and capable of rapidly analyzing and predicting flows of thousands of pedestrians. The proposed framework is one of the first steps towards fully integrated CCTV-based automated pedestrian traffic management system.

A copy of Muhammad’s dissertation is available for examination from Karen Underwood, Department of Computational and Data Sciences, 373 Research Hall. The dissertation is available to read only within the Department and cannot be taken out of the Department or copied.

Notice and Invitation
Oral Defense of Doctoral Dissertation
Doctor of Philosophy in Computational Social Science
Department of Computational and Data Sciences
College of Science
George Mason University

Karl Selke
Bachelor of Science, Lake Superior State University, 2004
Master of Science, George Washington University, 2006

REALLAND: A WARGAMING APPROACH TO
COMPUTATIONAL INTERNATIONAL RELATIONS

Tuesday, December 5, 1:00 p.m.
Research Hall, Room 92

All are invited to attend.

Committee
Claudio Cioffi-Revilla, Dissertation Director
William Kennedy
Margaret Polski
Peter Perla

The objective of this study is to advance a wargaming approach to computational international relations (IR). Wargaming has a long tradition of artfully balancing between realism and human playability. Those same techniques can be incorporated into computational IR models. A wargaming approach affords both greater resolution in the operational environment and emulation of human decision-making, which provides a significant alternative to past computational IR models of international conflict. As a demonstration of this approach, I developed two simulations: Basic and Enhanced RealLand.

Basic RealLand is a replication of the work of past researchers (Cusack and Stoll, 1990; Duffy, 1992) providing a comparative foundation for innovation within Enhanced RealLand. Enhanced RealLand combines computational social science (CSS) techniques with defense and commercial wargaming mechanisms to enable a wargaming approach to computational IR. It is a strategic-operational simulation where players sense the world, identify issues, develop strategies, and implement actions such as trade, alliance building, and war. The shift away from game-theoretic approaches for modeling computer agents as nation-states to the conflict-theory model as players proved promising, and the results generate a world worthy of the prominent IR realist theorists. An important contribution is the creation of a pseudo history, similar to that of a narrative resulting from a social simulation.

This study advanced a wargaming approach to computational IR by demonstrating that additional representative modeling, in a human playable form, can be used for advanced IR research.   By placing the focus on simulating human decision-making through a descriptive process, Enhanced RealLand provides an approach and extensible framework for computational IR models to have analytic utility whether for theorists, policy-makers, or educators.

A copy of Karl’s dissertation is available for examination from Karen Underwood, Department of Computational and Data Sciences, 373 Research Hall. The dissertation is available to read only within the Department and cannot be taken out of the Department or copied.

Notice and Invitation
Oral Defense of Doctoral Dissertation
Doctor of Philosophy in Computational Social Science
Department of Computational and Data Sciences
College of Science
George Mason University

Peter Revay
BNUS, Masaryk University, 2010
MS, University of Vermont, 2014

THE EFFECTS OF DUAL INHERITANCE MECHANISMS OF CULTURAL EVOLUTION ON EMERGENT SOCIAL NETWORKS

Thursday, December 7, 2:0 p.m.
Exploratory Hall, Room 162

All are invited to attend.

Committee
Dr. Claudio Cioffi, Dissertation Chair
Dr. Andrew Crooks
Dr. Sean Luke
Dr. J. Daniel Rogers

Understanding cultural dynamics in human societies is the first step towards solving many complex social issues. In this dissertation I focus on the drivers of diffusion and adoption of cultural traits, such as values, beliefs, and behaviors. I adopt an evolutionary view of cultural dynamics. Particularly, I use concepts from dual inheritance theories of cultural evolution to develop and test an agent-based model capable of simulating the changing distributions of cultural traits in a large population of actors over the course of prolonged periods of time. Particularly, I pay close attention to the mechanisms of indirectly biased transmission of traits and guided variation, which are both hypothesized to be significant drivers of cultural dynamics. Indirectly biased transmission consists of the adoption of specific trait variants on the basis of possession of initially unrelated external markers. Guided variation is then individual adaptation driven by self-exploration.

I use various methods to explore the pathways of cultural evolution. Among them are agent-based modeling, evolutionary computation, complex network analysis and statistical data analysis. Furthermore, I make use of large publicly available datasets to validate my models. The first one of these is the database of bill co-sponsorship in the U.S. House of Representatives from 1973 to 2008. The other is a comprehensive dataset of scientific co-authorship in various disciplines stretching back for over a century. The results show that cultural evolution models based on indirectly biased transmission and guided variation are suitable to explaining the dynamics of various complex social networks. Furthermore, I show that this type of cultural evolution leads to emergence of meaningful cultural signs by gradually associating previously independent external markers with specific cultural trait variants. Finally, I describe how the proposed model leads to plausible social network configurations.

A copy of Peter’s dissertation is available for examination from Karen Underwood, Department of Computational and Data Sciences, 373 Research Hall. The dissertation is available to read only within the Department and cannot be taken out of the Department or copied.

Notice and Invitation
Oral Defense of Doctoral Dissertation
Doctor of Philosophy in Computational Social Science
Department of Computational and Data Sciences
College of Science
George Mason University

Shawn Joseph Bucholtz
Bachelor of Science, Michigan State University, 2000
Master of Science, University of Maryland, 2003

THE WASHINGTON, D.C. HOUSING AFFORDABILITY SIMULATOR

Thursday, December 7, 9:30 a.m.
Exploratory Hall, Room 162

All are invited to attend.

Committee
Dr. Robert Axtell, Dissertation Director
Dr. Andrew Crooks
Dr. William Kennedy
Dr. Katrin Anacker

This dissertation presents the Washington, D.C. Housing Affordability Simulator, or DCHAS. DCHAS is an empirical agent-based model of urban housing supply and demand, with a special emphasis on housing affordability and affordable housing production. DCHAS agents include households, landlords, developers and the local government. Past agent-based and microsimulation modeling efforts have demonstrated the importance of including agent heterogeneity and land markets in models of urban housing supply and demand. DCHAS builds upon this foundation and extends prior efforts by including six additional features important to on housing affordability and affordable housing production: agent variation appropriate to low-income households, explicit representation of Federal housing subsidies, explicit representation of affordable housing supply, rental control, zoning and regeneration of properties, and filtering and rehabilitation of housing units.

DCHAS is calibrated to the population and housing stock as it existed in 2010. The behaviors of DCHAS’s agent are parameterized with data from 2011 to 2015. Combining a 2010-based population and housing stock with agent behavior parameterized with data from 2011 to 2015, it is demonstrated that DCHAS reliably reproduces housing supply and demand outcomes observed in 2015. Then, DCHAS is used to simulate four housing supply and demand scenarios over the next ten years (2016 -2025). The principle contributions of this dissertation are to: (1) identify and explore concepts critical to housing affordability in an urban environment; (2) demonstrate how to empirically represent these concepts through the use of administrative data sources, and (3) demonstrate how to build an empirically-based ABM that can be used to simulate housing affordability under different market conditions or housing policy scenarios.

A copy of Shawn’s dissertation is available for examination from Karen Underwood, Department of Computational and Data Sciences, 373 Research Hall. The dissertation is available to read only within the Department and cannot be taken out of the Department or copied.

Notice and Invitation
Oral Defense of Doctoral Dissertation
Doctor of Philosophy in Computational Sciences and Informatics
Department of Computational and Data Sciences
College of Science
George Mason University

Joel M. Mejeur
Bachelor of Science, Taylor University, 2000
Master of Science, University of Tennessee, 2002
Master of Engineering-Physics, University of Virginia, 2008

FUNCTION SPACE NONLINEAR RESCALING METHODS FOR ELLIPTIC
CONTROL PROBLEMS WITH POINT-WISE STATE AND CONTROL CONSTRAINTS

Thursday, November 30, 10:30 a.m.
Exploratory Hall, Room 3301

All are invited to attend.

Committee
Igor Griva, Dissertation Director
Padmanabhan Seshaiyer
Chi Yang
Fernando Camelli

State inequality constraints in PDE Constrained Optimization (PDECO) arise in many areas of science and engineering. Unfortunately these constraints, and the resulting Lagrange multipliers, are known to negatively influence the behavior of many existing optimization methods. In this work Nonlinear Rescaling based methods are used for the state and control constraints. In particular, a Nonlinear Rescaling-Primal Dual Augmented Lagrangian method is analyzed and proven to have linear convergence for state and control constrained problems. In addition, a Primal Dual Nonlinear

Rescaling Augmented Lagrangian method is analyzed for control constraints and shown to have superlinear convergence properties. In each of the derived methods the Finite Element Method will be used to construct and solve the discretized version of the inner iteration.

A copy of Joel’s dissertation is available for examination from Karen Underwood, Department of Computational and Data Sciences, 373 Research Hall. The dissertation is available to read only within the Department and cannot be taken out of the Department or copied.

Notice and Invitation
Oral Defense of Doctoral Dissertation
Doctor of Philosophy in Computational Sciences and Informatics
Department of Computational and Data Sciences
College of Science
George Mason University

Daniel Ray Sponseller
Bachelor of Science, Virginia Polytechnic Institute and State University, 1987
Master of Science, Virginia Polytechnic Institute and State University, 1995

Molecular Dynamics Study of Polymers and Atomic Clusters

Thursday, November 30, 10:00 a.m.
Research Hall, Room 161

All are invited to attend.

Committee
Estela Blaisten-Barojas, Dissertation Director
Howard Sheng
James Glasbrenner
Dmitri Klimov

This dissertation contains investigations based on Molecular Dynamics (MD) of a variety of systems, from small atomic clusters to polymers in solution and in their condensed phases. The overall research is divided in three parts. First, I tested a new thermostat in the literature on the thermal equilibration of a small cluster of Lennard-Jones (LJ) atoms. The proposed thermostat is a Hamiltonian thermostat based on a logarithmic oscillator with the outstanding property that the mean value of its kinetic energy is constant independent of the mass and energy. I inspected several weak-coupling interaction models between the LJ cluster and the logarithmic oscillator in 3D. In all cases I show that this coupling gives rise to a a kinetic motion of the cluster center of mass without transferring kinetic energy to the interatomic vibrations. This is a failure of the published thermostat because the temperature of the cluster is mainly due to vibrations in small atomic clusters This logarithmic oscillator cannot be used to thermostat any atomic or molecular system, small or large.

The second part of the dissertation is the investigation of the inherent structure of the polymer polyethylene glycol (PEG) solvated in three different solvents: water, water with 4% ethanol, and ethyl acetate. PEG with molecular weight of 2000 Da (PEG2000) is a polymer with many applications from industrial manufacturing to medicine that in bulk is a paste. However, its structure in very dilute solutions deserved a thorough study, important for the onset of aggregation with other polymer chains. I introduced a modification to the GROMOS 54A7 force field parameters for modeling PEG2000 and ethyl acetate. Both force fields are new and have now been incorporated into the database of known residues in the molecular dynamics package Gromacs. This research required numerous high performance computing MD simulations in the ARGO cluster of GMU for systems with about 100,000 solvent molecules. My findings show that PEG2000 in water acquires a ball-like structure without encapsulating solvent molecules. In addition, no hydrogen bonds were formed. In water with 4% ethanol, PEG2000 acquires also a ball-like structure but the polymer ends fluctuate folding outward and onward, although the general shape is still a compact ball-like structure.

In contrast, PEG2000 in ethyl acetate is quite elongated, as a very flexible spaghetti that forms kinks that unfold to give rise to folds and kinks in other positions along the polymer length. The behavior resembles an ideal polymer in a theta solvent. A Principal Component Analysis (PCA) of the minima composing the inherent structure evidences the presence of two distinct groups of ball-like structures of PEG2000 in water and water with 4\% ethanol. These groups give a definite signature to the solvated structure of PEG2000 in these two solvents. In contrast, PCA reveals several groups of avoided states for PEG2000 in ethyl acetate that disqualify the possibility of being an ideal polymer in a theta solvent.

The third part of the dissertation is a work in progress, where I investigate the condensed phase of PEG2000 and study the interface between the condensed phase and the three different solvents under study. With a strategy of combining NPT MD simulations at different temperatures and pressures, PEG2000 condensed phase displays the experimental density within a 1% discrepancy at 300 K and 1 atm. This is a very encouraging result on this ongoing project.

A copy of Dan’s dissertation is available for examination from Karen Underwood, Department of Computational and Data Sciences, 373 Research Hall. The dissertation is available to read only within the Department and cannot be taken out of the Department or copied.

Notice and Invitation
Oral Defense of Doctoral Dissertation
Doctor of Philosophy in Computational Social Science
Department of Computational and Data Sciences
College of Science
George Mason University

Elaine Reed
Bachelor of Science, Virginia Polytechnic Institute and State University, 1985
Master of Science, George Washington University, 1989

The Emergence of Self-government:
Agentizing Acemoglu and Robinson

Thursday, November 30, 1:30 p.m.
Research Hall, Room 91

All are invited to attend.

Committee
Robert Axtell, Dissertation Director
Andrew Crooks
Richard Wagner

Institutions are created by people interacting in complex ways with others in their socio-economic environment. A study of institutions should therefore study the people and interactions that create them. Acemoglu and Robinson (hereafter A&R) developed a game-theoretic framework for analyzing how economic incentives influence the way social groups shape institutions to allocate political and economic power. The A&R models assume groups or classes of people are completely rational and identical intra-group. Analytical difficulties impede A&R from exploring more realistic interactions. This dissertation utilizes an agent-based computational methodology to represent a subset of the A&R models. The computational model permits agents to be heterogeneous, which can affect outcomes at the group and aggregate levels. Specifically, with intra-group homogeneity the agent-based model reproduces the group-level threshold conditions affecting institutional choices found by A&R. I show that these results are robust to parameter changes within the ranges defined by A&R. However, I then relax the intra-group homogeneity assumption, allowing exploration of the roles of income distributions and bounded intelligence on the larger outcomes for all groups. The population structure with heterogeneity can include a more realistic middle class. Modeling a middle class by using agent-based models with heterogeneous agents finds that the effect of a middle class is non-linear. This dissertation demonstrates the usefulness of agent-based modeling as a viable alternative quantitative methodology for studying complex institutions.

A copy of Elaine’s dissertation is available for examination from Karen Underwood, Department of Computational and Data Sciences, 373 Research Hall. The dissertation is available to read only within the Department and cannot be taken out of the Department or copied.

Notice and Invitation
Oral Defense of Doctoral Dissertation
Doctor of Philosophy in Computational Sciences and Informatics
Department of Computational and Data Sciences
College of Science
George Mason University

Jeffrey Stevens
Bachelor of Science, State University of New York at Buffalo, 1992
Master of Science, State University of New York at Buffalo, 1994

New Methods Of Spectral-Density Based Graph Construction And Their Application To Hyperspectral Image Analysis

Monday, November 20, 2017, 2:00 -4:00 p.m.
Research Hall, Room 162

All are invited to attend.

Committee
Dr. John J. Qu, Dissertation Director
Dr. Ronald G. Resmini
Dr. Robert S. Weigel
Dr. Juan R. Cebral

The past decade has seen the emergence of many hyperspectral image (HSI) analysis algorithms based on graph theory and derived manifold-coordinates. Yet, despite the growing number of algorithms, there has been limited study of the graphs constructed from spectral data themselves. Which graphs are appropriate for various HSI analyses—and why? This research aims to begin addressing these questions as the performance of graph-based techniques is inextricably tied to the graphical model constructed from the spectral data.

We advance the performance of all graph- and manifold-based analyses by providing guidelines developed from rigorous study to ensure the starting graphs are of higher quality, and demonstrate the effectiveness of shared nearest neighbors (SNN) and mutual proximity (MP) as alternative methods of graph construction not limited by assumptions of current techniques. In developing construction guidance, we consider the impacts of generating graphs in higher dimensional HSI data and present the first study of its kind related to concentration of measure and hubness in graphs of HSI data indicating the types of analyses that can be impacted by these phenomena. To conclude, we introduce new algorithms based on SNN and MP for classification and anomaly detection using a Schroedinger Eigenmaps formulation and compare performance to leading techniques.

A copy of Jeff’s dissertation is available for examination from Karen Underwood, Computational and Data Sciences, 373 Research Hall. The dissertation is available to read only within the Department and cannot be taken out of the department or copied.

Notice and Invitation
Oral Defense of Doctoral Dissertation
Doctor of Philosophy in Computational Sciences and Informatics
Department of Computational and Data Sciences
College of Science
George Mason University

Daniel Pulido
Bachelor of Science, Boston University, 1998
Master of Science, Worcester Polytechnic Institute, 2003

Self-Similar Spin Images for Point Cloud Matching

Friday, October 13, 2017, 10:00 a.m.
Research Hall, Room 162

All are invited to attend.

Committee
Anthony Stefanidis, Dissertation Director
Estela Blaisten-Barojas
Arie Croitoru
Juan Cebral

The rapid growth of Light Detection And Ranging (Lidar) technologies that collect, process, and disseminate 3D point clouds have allowed for increasingly accurate spatial modeling and analysis of the real world. Lidar sensors can generate massive 3D point clouds of a collection area that provide highly detailed spatial and radiometric information. Simultaneously, the growth of other forms of geospatial data (e.g., crowdsourced Web 2.0 data) have provided researchers with a wealth of freely available data sources that cover a variety of geographic areas. However, combining data from disparate sources requires overcoming numerous technical challenges in order to generate products that mitigate their respective disadvantages and combine their advantages.

Therefore, this dissertation addresses the problem of fusing two point clouds from potentially different sources by considering two specific problems: scale matching and feature matching. To address the problem of feature matching we develop a novel feature descriptor referred to as “Self-Similar Spin Images” which combine the concept of local self-similarity with the descriptive power of Spin Images. To address the problem of scale matching we develop a novel scale detection metric referred to as “Self-Similar Keyscale” which analyzes the self-similarity of two point clouds to identify a characteristic scale to match them. Finally, we develop a novel change detection method as a sample use case of the developed matching techniques.

Notice and Invitation
Oral Defense of Doctoral Dissertation
Computational Social Science
Department of Computational and Data Sciences
College of Science
George Mason University

JONATHAN GOLDSTEIN
Bachelor of Arts, Cornell University, 2004

Rethinking Housing with Agent-based Models: Models of the Housing Bubble and Crash in the Washington, D.C. Area 1997-2009

Monday, April 24, 2017, 10:00-Noon
Research Hall, Room 92

All are invited to attend.

Committee
Robert Axtell, Chair
William G. Kennedy
Carlos Ramirez
J. Doyne Farmer

This dissertation presents a series of related agent-based models (ABMs) of the housing market of the Washington DC Metropolitan Statistical Area. The models investigate the causes of the housing market bubble and crash during the time period 1997-2009 and policies that could have avoided such a crisis. The work in this dissertation contributed to three research areas: understanding the underlying causes of the housing crisis, demonstrating the utility of ABMs for investigating macro phenomena, and improving ABM methodology.

Using the housing market models, I ran counterfactual analyses to investigate causes and potential mitigations of the crisis. I show that leverage and expectations are the two most prominent causes of the bubble, but that other factors, such as interest rates, the norm governing the share of income to spend on housing, and seller behavior influence the bubble. I find that lending standards and refinance rules play almost no part in the bubble, contrary to some theories of the housing crisis. Towards the end of the dissertation, I pair the housing market with a model of mortgage-backed securities. I show that the increased velocity of lending made possible by securitization can increase the size of bubbles and make markets more fragile, increasing the likelihood of crashes.

The ABMs in this dissertation exploit multiple large, heterogeneous data sets and utilize realistic behavioral rules to match detailed housing market dynamics. Input data include loan level data, multiple listing service records, and demographic information from a variety of sources. The ABMs exploit this data by choosing the precise areas of input distributions to use based on the context of the model. This allows the ABMs to match not only aggregate outputs, but intermediate outputs and data distributions. For example, the ABMs in this dissertation not only reproduce empirical macro phenomena, such as the shape of the house price index, but also intermediate variables (e.g., distribution of loan types, average leverage, average days on market, average ratio of sold price to original listing price) and output distributions (e.g., distribution of house prices).

Throughout the dissertation, I follow several methodological principles in construction and analysis of the ABMs. First, I demonstrate the use of data to constrain the models as much as possible. Next, I describe a sensitivity analysis methodology that goes beyond parametric variations, but also varies model rules in what I term a structural sensitivity analysis. I demonstrate how criticisms about ABMs with regard to their opacity, brittleness, and dependency on arbitrary modeling decisions can be resolved through such an analysis. I also describe the architectural design of the models, which makes explicit the theoretically-inspired behavioral rules, facilitating structural sensitivity analyses.

A copy of Jon’s dissertation is available for examination from Karen Underwood, Department of Computational and Data Sciences, 373 Research Hall. The dissertation is available to read only within the Department and cannot be taken out of the Department or copied.