Horizon dependence of utility optimizers in incomplete models

Yu, Hang Carnegie Mellon University 2011 해외박사(DDOD)

The utility maximization of terminal wealth in continuous time has a long history. Literally hundreds of papers have studied many facets of such questions. Despite of the vast amount of literature, one aspect seems not well understood to date. That is the stability of the optimizers. To the best of my knowledge, only a few papers have been published on this topic, such as [3] and [21]. This document contains a paper that I participated in during my PhD research at Carnegie Mellon University. In this paper we studied the stability problem in time horizon of utility maximization in incomplete models. The question we were interested in was how the planning horizon affected the optimal investment decision. A shorter version of this thesis has been accepted for publication (jointly with my advisor Kasper Larsen) in Finance & Stochastics . The main contribution of this work is that we identified the models that fail to be stable and we also provided conditions preventing the existence of this phenomenon.

Photoproduction of eta and eta' mesons using CLAS at JLAB

Krahn, Zebulun A Carnegie Mellon University 2007 해외박사(DDOD)

This work represents the first Partial Wave Analysis(PWA) to emerge from the PWA program at Carnegie Mellon University. The focus of this particular analysis is photoproduced eta and eta' mesons. The data were obtained in Hall B at Jefferson Lab using a photon beam at an endpoint energy of 4.016 GeV in the spring and summer of 2004. The Hall B photon tagging system and the CLAS spectrometer were used to collect the data. From this data, and in the course of performing a partial wave analysis of the Neta and Neta' systems, measurements were made of the differential cross sections for eta and eta' photoproduction. The differential cross section measurements for the eta improve the statistics and slightly improve the kinematic range for world data. The results for the eta' represents an almost entirely new dataset which will allow for further analysis of the eta' meson. The results of this measurement for the eta' extend the statistics and kinematic range of world data dramatically.

Essays in education economics

Davis, Billie S Carnegie Mellon University 2013 해외박사(DDOD)

My dissertation examines the causes of, and one possible program to help alleviate, urban public school district enrollment decline. Enrollment decline has plighted many cities in the Northeast and Midwest over the past two decades, leading to financial struggles and, sometimes, large school closures or reorganizations. In Chapter 1, I describe how changing birth rates and regional migration have contributed to a decline in the number of school-age children in the Northeast and Midwest (NEMW) census regions of the United States. In Chapter 2, "Determinants of Urban Enrollment Decline", I show that competitive pressures from suburban, charter, and private schools have caused disproportionate enrollment losses in large urban public school districts. Specifically, I use a panel of data on public, charter, and private school enrollment and city characteristics to analyze the impact of metropolitan enrollment losses and substitution to charter and private schools on the urban core district's education "market share". Finally, in Chapter 31, "Bounding the Retention Effects of a Gifted Program Using a Modified Regression Discontinuity Design", we examine whether gifted programs can help urban districts retain students with higher ES backgrounds. Gifted programs often employ IQ thresholds for admission, with those above the threshold being admitted. These types of admission rules are often mandated by state rules and create strong incentives D manipulate the IQ score of students to increase access to the program. We propose two new tests that can be used to detect local manipulation of IQ scores. In the presence of local manipulation, the standard regression discontinuity estimator does not identify the local average treatment effect of the program. We show how to codify the approach to construct a lower bound for the effectiveness of the program. This lower bound can be estimated using a modified RD estimator. Our application uses a new and unique data set that is based on applications and admissions to the gifted program of an anonymous urban school district. Our point estimates suggest that there is a favorable effect on retention for students in higher SES households. 1Co-authored with John Engberg, a senior researcher at the RAND Corporation; Dennis Epple, the Thomas Lord Professor at Carnegie Mellon University; Holger Sieg, the J.M. Cohen Term Professor of Economics at the University of Pennsylvania; and Ron Zimmer, Associate Professor of Public Policy and Education at Vanderbilt University.

Micromagnetic design, analysis, and optimization of 1 Tbit/square inch MFM-based scanning-probe storage device

El-Sayed, Rany Tawfik Carnegie Mellon University 2004 해외박사(DDOD)

In this research we have extensively analyzed and optimized the performance of MEMS-positioned, nonvolatile, re-writable, mass storage devices with MFM-based single-pole read/write magnetic tips currently under development at Carnegie Mellon University. This has been achieved through general design consideration investigation, analytical modeling, and through in-depth micromagnetic modeling and simulations. For micromagnetic modeling, two simulators were especially designed and built by the author to study and analyze the performance of the MFM-based MEMS-actuated mass storage device. The simulators have been tested and verified using the standard problem #1 proposed by the Micromagnetic Modeling Activity Group (muMAG) in the NIST Center for Theoretical and Computational Materials Science (CTCMS). In general design consideration investigation, many important concepts and observations have been developed. First, the tip field amplitude and gradient dependence on tip size has been studied. We then investigated MFM and deduced the concept of the "critical ring" which determines which ring of grains are the most influential on the overall force. We also investigated the equivalent demagnetization fields and introduced a "self-magnetization limit" guideline to be used in designing the most efficient media. We concluded the general design consideration analysis by studying the interaction between an ideal system of a tip and a media. By studying the optimum recording process through tip field gradient optimization, we were able to deduce a framework that relates the tip geometry with the media properties to produce the optimum theoretical SNR. Performance optimization for the single-layer media case has been extensively studied. Using the impulse-response and inverse-convolution technique, we deduced a novel tip trajectory for the optimum recording process. Detection/Recording heights that gives the best performance in terms of areal density and SNR were deduced. The medium structure is then optimized in terms of maximizing the SNR resulting in two structures namely, Co76Cr20Pt 4 and Co83Cr10Pt7. Four design points where deduced from these configurations showing areal densities of approximately 0.3 Tb/in2 with an overall SNR of 20--25 dB assuming 12 pN equivalent electronic noise. Dynamic monitoring of media reversal has revealed "ring effect" formation due to high demagnetization fields relative to the tip field. (Abstract shortened by UMI.).

Measuring and Modeling Security and Privacy Laws

Romanosky, Sasha Carnegie Mellon University 2012 해외박사(DDOD)

This manuscript presents empirical and analytical analysis and discussion of security and privacy laws. The introduction, together with the three substantive chapters each represent separate research papers written as partial fulfillment of my PhD dissertation in the Heinz College, Carnegie Mellon University. Chapter 2 is an abbreviated version of a paper coauthored with Alessandro Acquisti and published in the Berkeley Technology Law Journal (Romanosky and Acquisti 2009). The full paper examines three alternative policy interventions that can be applied to reduce the externalities caused by data breaches. Moreover, it examines the privacy costs through the lens of legal and economic theory. I would like to thank John Bagby, Fred Cate, Ben Edelman, Mark Melodia, and Alana Maurushat for their insightful comments and feedback, and Charlotte Chang, Varty Defterderian, Kristin Kemnitzer, and Peter Nagle for their editing. Chapter 3 empirically estimates the effect of data breach disclosure laws on identity theft. The paper from which it was drawn was coauthored with my advisors Rahul Telang and Alessandro Acquisti and was published in the Journal of Policy Analysis and Management (JPAM)1. I would like to thank Katrina Baum, Al Blumstein, Laura Dugan, Vasundhara Garg, John Hutchins, Jed Kolko, Thad Kousser, Anand Nandkumar, JJ Prescott, Peter Swire, and Ellerie Webber for their valuable suggestions. Rahul Telang acknowledges generous support of NSF (National Science Foundation) through the CAREER award grant CNS-0546009. Chapter 4 analytically examines disclosure laws and addresses the conditions under which mandatory disclosure could reduce social costs. It was coauthored with my advisor, Alessandro Acquisti, and Richard Sharp, a mathematician who authored the mathematical proofs. I would like to thank Nicolas Christin and participants of the Ninth Workshop on the Economics of Information Security for their comments and suggestions. Chapter 5 empirically examines US civil lawsuits relating to the loss or theft of personal information, and was coauthored with Alessandro Acquisti and David Hoffman of the Beasley School of Law at Temple University. This research was supported by Temple Law School's Conwell Corps Program. We would like to thank Antima Chakraborty, Carol Anne Donohoe, Ian Everhart, Caitlin Jones, Kevin Leary and Jake Oresick for their research assistance. We would also like to thank Paul Bond, Aaron Burnstein, Fainna Kagan, Amelia Haviland, Mark Melodia, Kristen Matthews, Peter Oh, Barrie Nault, David Navetta, Mohammad Rahman, Theresa Romanosky, Boris Segalis, Brendon Tavelli and 7 anonymous attorneys for their valuable insights and suggestions. Finally, I would like to acknowledge CyLab at Carnegie Mellon for their generous support for this research under grants DAAD19-02-1-0389 and W911NF-09-1-0273 from the Army Research Office. 1Romanosky, S., Telang, R. & Acquisti, A. (2011). Do Data Breach Disclosure Laws Reduce Identity Theft? Journal of Policy Analysis and Management, 30(2), 256-286. Available at http://papers.ssrn.com/sol3/papers.cfm?abstract_id=1268926.

Adaptive binary search trees

Derryberry, Jonathan Carlyle Carnegie Mellon University 2009 해외박사(DDOD)

A ubiquitous problem in the field of algorithms and data structures is that of searching for an element from an ordered universe. The simple yet powerful binary search tree (BST) model provides a rich family of solutions to this problem. Although BSTs require O(lg n) time per operation in the worst case, various adaptive BST algorithms are capable of exploiting patterns in the sequence of queries to achieve tighter, input-sensitive, bounds that can be o(lg n) in many cases. This thesis furthers our understanding of what is achievable in the BST model along two directions. First, we make progress in improving instance-specific lower bounds in the BST model. In particular, we introduce a framework for generating lower bounds on the cost that any BST algorithm must pay to execute a query sequence, and we show that this framework generalizes previous lower bounds. This suggests that the optimal lower bound in the framework is a good candidate for being tight to within a constant factor of the optimal BST algorithm for each input. Further, we show that lower bounds in this framework are also valid lower bounds for instances of the partial-sums problem in a restricted model of computation, which suggests that augmented BSTs may be the most efficient way of maintaining sums over ranges of an array when the entries of the array can be updated throughout time. Second, we improve the input-sensitive upper bounds that are known to be achievable in the BST model by introducing two new BST algorithms, skip-splay and cache-splay. These two algorithms are the first BSTs that are known to have running times that have nontrivial competitiveness to Iacono's Unified Bound, which is a generalization of the dynamic finger and working set bounds. Skip-splay is a simple algorithm that is nearly identical to splaying, and it achieves a running time that is within additive O(lg lg n) per operation of the Unified Bound. Cache-splay is a slightly more complicated splay-based algorithm that is the first BST to achieve the Unified Bound to within a constant factor.

Experimental analysis on the mechanical properties of vitrified blood vessels

Jimenez Rios, Jorge L Carnegie Mellon University 2007 해외박사(DDOD)

Ice formation is widely accepted as the cornerstone of cryoinjury, which hinders the widespread application of cryopreservation of biological tissue. Vitrification is an ice-free alternative to conventional cryopreservation, achieved by introducing high concentrations of cryoprotective agents (CPAs) to the tissue. The CPA interacts strongly with water, preventing it from crystallizing while it is cooled below its nucleation point, leading to a supercooled liquid. CPA viscosity increases with decreasing temperature until the tissue becomes trapped in a glass-like state (vitreous is Latin for glassy), and remains in this form without appreciable degradation. This technique results in a marked improvement over traditional cryopreservation in terms of post-thawing functionality of the material. Vitrification has been successfully applied to small biological specimens. However, its application to large tissue samples has been hampered by mechanical damage to the material. The high cooling rates needed for these processes create large thermal gradients, which result in mechanical stresses driven by the thermal expansion of the tissue. It is theorized that these mechanical stresses are the main cause of tissue injury during vitrification, as they can surpass the strength of the material. While the importance of thermo-mechanical stresses to the post-thawing viability and functionality of vitrified tissue has been acknowledged, a comprehensive engineering analysis of this problem has not been attempted until recently, by the Biothermal Technology Laboratory at Carnegie Mellon University. A continuum mechanics model of the vitrified tissue could assist in predicting these stresses, for the purpose of developing better methods of storage, subject to lower stresses. This dissertation presents an experimental study which measures the key mechanical properties of vitrified tissue for subsequent use in mechanical modeling. An experimental system was designed and constructed to mimic typical vitrification protocol on a blood vessel model. Results are presented for thermal expansion, shear modulus, and viscosity in the cryogenic temperature range, showing viscoelastic behavior above glass transition and linear-elastic behavior below it. The tissue models are carotid arteries from goats and calves permeated with the cryoprotectant cocktails VS55 and DP6 and the reference solution of 7.05M DMSO.

Statistical inference in mapping and localization for mobile robots

Araneda, Anita Carnegie Mellon University 2004 해외박사(DDOD)

A fundamental capability of a truly autonomous mobile robot is its ability to build a map of its environment using the data gathered during navigation. The problem of learning how to build such a map is usually called the Simultaneous Localization and Mapping (SLAM) problem, as the locations visited by the robot need to be known in order for it to build a map. This thesis develops a statistical solution to the SLAM problem. We formalize the SLAM problem using a Graphical Representation. The data correspond to the locations visited by the robot, obtained from a noisy odometer, and the distances to the closest obstacles from each of those locations, obtained from a noisy laser sensor. The map corresponds to an occupancy grid. We introduce the set of true locations visited by the robot, and true distances to obstacles, as latent variables. Our formulation of the problem leads naturally to the estimation of the posterior distribution of maps given the data. We exploit particular factorizations of this distribution which allow us to implement three versions of importance sampling. We present the results obtained by these algorithms when applied to a data set obtained by a robot navigating inside Wean Hall Building at Carnegie Mellon University. One of our algorithms obtains successful results, and it emphasizes the need of including laser information in order to correct odometry error. (Abstract shortened by UMI.).

Emergent architectures: A case study for outdoor mobile robots

Gowdy, Jay Willard Carnegie-Mellon University 2000 해외박사(DDOD)

Software reuse is a key issue in any long term software engineering endeavor, such as the ongoing development of robotics systems. Existing approaches to software reuse involve fixing part of the software landscape as a constant foundation to build upon. Some approaches fix the software architecture, i.e., how data flows through the system, and then allow the user to swap components in and out. Other approaches define a suite of stable components and interfaces which can be mixed, matched, and extended. All of these approaches assume that the key to software reuse is pervasive and long lasting standards. Unfortunately, in a young domain such as mobile robotics both software components and software architectures are items of research, and thus are in constant flux. This dissertation proposes the replacement of global, system-wide, permanent standards with local, transient standards in the form of reconfigurable interfaces. These interfaces are not simply libraries of system calls, but contain mediators and adapters which stand between a software module and the system architecture, providing a module with a locally stable view of the system in which it resides, translating the module's requests and notifications into the protocols and data flow of that particular architecture. This system of mediation facilitates the emergence of a system architecture from the current needs and requirements of the tasks, rather than forcing the a priori adoption of any single architecture. To demonstrate the feasibility of this approach, two very different robotic tasks were examined: first the implementation of a dedicated, high-speed road following system under Carnegie Mellon's NavLab project, and second the implementation of a multi-modal military reconnaissance vehicle for the Unmanned Ground Vehicle program. Radically different architectures and components emerge from the requirements of these two tasks and the current capabilities of the component modules, but it is demonstrated that reconfigurable interfaces allow the moving of critical code between the two without even recompilation. The mediating abilities of reconfigurable interfaces are provided without requiring any pervasive, long-lasting communications standards while only requiring the execution overhead of a few pointer dereferences. The emergent architecture approach acknowledges the inevitable fluidity of both robotic components and system architectures. The reconfigurable interfaces act as a buffer for change flowing from the bottom up or change flowing from the top down, thus allowing the concurrent development of modules and architectures. Reconfigurable interfaces enable a natural, incremental, and empirical approach to building robotic systems in which the ultimate system architecture is free to emerge from the current task requirements and component capabilities, and the component developers are free to make improvements in their modules without an overriding regard to the systems in which they reside.

A planar robot for high-performance manipulation

Quaid, Arthur E Carnegie-Mellon University 2000 해외박사(DDOD)

A novel planar robot is used to demonstrate that large numbers of axes are not required for effective manipulation, and that there can be strong advantages to eliminating excess degrees-of-freedom. This robot combines sub-micron precision with a meter-sized workspace, motions with velocities above 1 m/s, and accelerations several times that of gravity. It combines planar stepping motor (Sawyer motor) actuation technology with a compact, high-precision, high-bandwidth position sensor developed at Carnegie Mellon University. The actuator operation is experimentally characterized over the closed-loop operating regime using an automated experimental setup that incorporates a laser interferometer and load cell. Parametric models are constructed based on these experiments that allow autonomous calibration of the actuators and sensors. This calibration enables on-site calibration, improving the precision of both positioning and force output without requiring external measuring devices. Coupled with software-based calibrated models and commutation, the stepping motor actuators can be operated as servo motors, simplifying the controller design and eliminating the underdamped oscillation modes that limit the robustness of open-loop stepping operation. PD and PID controllers are used to demonstrate the performance improvements from closed-loop operation, which include improved repeatability (to sub-micron levels), fast settling times (20 ms), and a four-fold decrease in energy usage. The main application of interest for this robot is conveyance and fine motion control of subproducts in a rapidly-deployable assembly system called minifactory. This system requires low-DOF robotic agents to cooperatively perform higher-DOF tasks, coordinating their actions through optical endpoint sensing, networked communications, or the task dynamics. It also requires each robotic agent to be trustworthy to ensure system reliability. The performance and robustness of closed-loop operation is demonstrated with a high-speed wall-following task, a high-precision visually guided positioning task, and a dynamic visual tracking task. A mobile parts feeder is also presented. This application exploits the dynamic range of the planar robot, which allows for gross motion and precision positioning of the feeder as well as a high enough bandwidth to generate 30 Hz vibratory waveforms. This application makes good use of the limited travel rotational axis of the robot.