UMD Theses and Dissertations
Permanent URI for this collectionhttp://hdl.handle.net/1903/3
New submissions to the thesis/dissertation collections are added automatically as they are received from the Graduate School. Currently, the Graduate School deposits all theses and dissertations from a given semester after the official graduation date. This means that there may be up to a 4 month delay in the appearance of a given thesis/dissertation in DRUM.
More information is available at Theses and Dissertations at University of Maryland Libraries.
Browse
6 results
Search Results
Item Copula Based Population Synthesis and Big Data Driven Performance Measurement(2019) Kaushik, Kartik; Cirillo, Cinzia; Civil Engineering; Digital Repository at the University of Maryland; University of Maryland (College Park, Md.)Transportation agencies all over the country are facing fiscal shortages due to the increasing costs of management and maintenance of facilities. The political reluctance to increase gas taxes, the primary source of revenue for many government transportation agencies, along with the improving fuel efficiency of automobiles sold to consumers, only exacerbate the financial dire straits. The adoption of electric vehicles threatens to completely stop the inflow of money into federal, state and regional agencies. Consequently, expansion of the network and infrastructure is slowly being replaced by a more proactive approach to managing the use of existing facilities. The required insights to manage the network more efficiently is also partly due to a massive increase in the type and volume of available data. These data are paving the way for network-wide Intelligent Transportation Systems (ITS), which promises to maximize utilization of current facilities. The waves of revolutions overtaking the usual business affairs of transportation agencies have prompted the development and application of various analytical tools, models and and procedures to transportation. Contributions to this growth of analysis techniques are documented in this dissertation. There are two main domains of transportation: demand and supply, which need to be simultaneously managed to effectively push towards optimal use of resources, facilities, and to minimize negative impacts like time wasted in delays, environmental pollution, and greenhouse gas emissions. The two domains are quite distinct and require specialized solutions to the problems. This dissertation documents the developed techniques in two sections, addressing the two domains of demand and supply. In the first section, a copula based approach is demonstrated to produce a reliable and accurate synthetic population which is essential to estimate the demand correctly. The second section deals with big data analytics using simple models and fast algorithms to produce results in real-time. The techniques developed target short-term traffic forecasting, linking of multiple disparate datasets to power niche analytics, and quickly computing accurate measures of highway network performance to inform decisions made by facility operators in real-time. The analyses presented in this dissertation target many core aspects of transportation science, and enable the shared goal of providing safe, efficient and equitable service to travelers. Synthetic population in transportation is used primarily to estimate transportation demand from Activity Based Modeling (ABM) framework containing well-fitted behavioral and choice models. It allows accurate verification of the impacts of policies on the travel behavior of people, enabling confident implementation of policies, like setting transit fares or tolls, designed for the common benefit of many. Further accurate demand models allow for resilient and resourceful planning of new or repurposing existing infrastructure and assets. On the other hand, short-term traffic speed predictions and speed based reliable performance measures are key in providing advanced ITS, like real-time route guidance, traveler awareness, and others, geared towards minimizing time, energy and resource wastage, and maximizing user satisfaction. Merging of datasets allow transfer of data such as traffic volumes and speeds between them, allowing computation of the global and network-wide impacts and externalities of transportation, like greenhouse gas emissions, time, energy and resources consumed and wasted in traffic jams, etc.Item SEQUENTIAL DECISION MAKING WITH LIMITED RESOURCES(2019) Sankararaman, Karthik Abinav; Srinivasan, Aravind; Computer Science; Digital Repository at the University of Maryland; University of Maryland (College Park, Md.)One of the goals of Artificial Intelligence (AI) is to enable multiple agents to interact, co-ordinate and compete with each other to realize various goals. Typically, this is achieved via a system which acts as a mediator to control the agents' behavior via incentives. Such systems are ubiquitous and include online systems for shopping (e.g., Amazon), ride-sharing (e.g., Uber, Lyft) and Internet labor markets (e.g., Mechanical Turk). The main algorithmic challenge in such systems is to ensure that they can operate under a variety of informational constraints such as uncertainty in the input, committing to actions based on partial information or being unaffected by noisy input. The mathematical framework used to study such systems are broadly called \emph{sequential decision making} problems where the algorithm does not receive the entire input at once; it obtains parts of the input by interacting (also called "actions") with the environment. In this thesis, we answer the question, under what informational constraints can we design efficient algorithms for sequential decision making problems. The first part of the thesis deals with the Online Matching problem. Here, the algorithm deals with two prominent constraints: uncertainty in the input and choice of actions being restricted by a combinatorial constraint. We design several new algorithms for many variants of this problem and provide provable guarantees. We also show their efficacy on the ride-share application using a real-world dataset. In the second part of the thesis, we consider the Multi-armed bandit problem with additional informational constraints. In this setting, the algorithm does not receive the entire input and needs to make decisions based on partial observations. Additionally, the set of possible actions is controlled by global resource constraints that bind across time. We design new algorithms for multiple variants of this problem that are worst-case optimal. We provide a general reduction framework to the classic multi-armed bandits problem without any constraints. We complement some of the results with preliminary numerical experiments.Item Branching diffusion processes in periodic media(2019) Hebbar, Pratima; Koralov, Leonid; Mathematics; Digital Repository at the University of Maryland; University of Maryland (College Park, Md.)In the first part of this manuscript, we investigate the asymptotic behavior of solutions to parabolic partial differential equations (PDEs) in $\real^d$ with space-periodic diffusion matrix, drift, and potential. The asymptotics is obtained up to linear in time distances from the support of the initial function. Using this asymptotics, we describe the behavior of branching diffusion processes in periodic media. For a super-critical branching process, we distinguish two types of behavior for the normalized number of particles in a bounded domain, depending on the distance of the domain from the region where the bulk of the particles is located. At distances that grow linearly in time, we observe intermittency (i.e., the $k-$th moment dominates the $k-$th power of the first moment for some $k$), while, at distances that grow sub-linearly in time, we show that all the moments converge. In the second part of the manuscript, we obtain asymptotic expansions for the distribution functions of continuous time stochastic processes with weakly dependent increments in the domain of large deviations. As a key example, we show that additive functionals of solutions of stochastic differential equations (SDEs) satisfying H\"ormander condition on a $d$--dimensional compact manifold admit asymptotic expansions of all orders in the domain of large deviations.Item Asymptotic problems for stochastic partial differential equations(2015) Salins, Michael; Cerrai, Sandra; Mathematics; Digital Repository at the University of Maryland; University of Maryland (College Park, Md.)Stochastic partial differential equations (SPDEs) can be used to model systems in a wide variety of fields including physics, chemistry, and engineering. The main SPDEs of interest in this dissertation are the semilinear stochastic wave equations which model the movement of a material with constant mass density that is exposed to both determinstic and random forcing. Cerrai and Freidlin have shown that on fixed time intervals, as the mass density of the material approaches zero, the solutions of the stochastic wave equation converge uniformly to the solutions of a stochastic heat equation, in probability. This is called the Smoluchowski-Kramers approximation. In Chapter 2, we investigate some of the multi-scale behaviors that these wave equations exhibit. In particular, we show that the Freidlin-Wentzell exit place and exit time asymptotics for the stochastic wave equation in the small noise regime can be approximated by the exit place and exit time asymptotics for the stochastic heat equation. We prove that the exit time and exit place asymptotics are characterized by quantities called quasipotentials and we prove that the quasipotentials converge. We then investigate the special case where the equation has a gradient structure and show that we can explicitly solve for the quasipotentials, and that the quasipotentials for the heat equation and wave equation are equal. In Chapter 3, we study the Smoluchowski-Kramers approximation in the case where the material is electrically charged and exposed to a magnetic field. Interestingly, if the system is frictionless, then the Smoluchowski-Kramers approximation does not hold. We prove that the Smoluchowski-Kramers approximation is valid for systems exposed to both a magnetic field and friction. Notably, we prove that the solutions to the second-order equations converge to the solutions of the first-order equation in an $L^p$ sense. This strengthens previous results where convergence was proved in probability.Item Mechanism and Chance: Toward an Account of Stochastic Mechanism for the Life Sciences(2014) DesAutels, Lane Thomas; Darden, Lindley; Philosophy; Digital Repository at the University of Maryland; University of Maryland (College Park, Md.)In this dissertation, my aim is to develop some important new resources for explaining probabilistic phenomena in the life sciences. In short, I undertake to articulate and defend a novel account of stochastic mechanism for grounding probabilistic generalizations in the life sciences. To do this, I first offer some brief remarks on the concept of mechanism in the history of philosophical thought. I then lay out some examples of probabilistic phenomena in biology for which an account of stochastic mechanism seems explanatorily necessary and useful: synaptic transmission in the brain, protein synthesis, DNA replication, evolution by natural selection, and Mendelian inheritance. Next, I carefully examine the concept of regularity as it applies to mechanisms--building on a recent taxonomy of the ways mechanisms may (or may not) be thought to behave regularly. I then employ this taxonomy to sort out a recent debate in the philosophy of biology: is natural selection regular enough to count as a mechanism? I argue that, by paying attention to the forgoing taxonomy, natural selection can be seen to meet the regularity requirement just fine. I then turn my attention to the question of how we should understand the chance we ascribe to stochastic mechanisms. To do this, I form a list of desiderata that any account of stochastic mechanism must meet. I then explore how mechanisms fit with several of the going philosophical accounts of chance: subjectivism, frequentism (both actual and hypothetical), Lewisian best-systems, and propensity. I argue that neither subjectivism, frequentism, nor best-system-style accounts of chance will meet all of the proposed desiderata, but some version of propensity theory can. Borrowing from recent propensity accounts of biological fitness and drift, I then go on to explore the prospects for developing a propensity interpretation of stochastic mechanism (PrISM) according to which propensities are (i) metaphysically analyzable and operationally quantifiable via a function of probability-weighted ways a mechanism might fire and (ii) not causally efficacious but nonetheless explanatorily useful. By appealing to recent analyses of deterministic and emergent chance, I argue further that this analysis need not be vulnerable to the threat of metaphysical determinism.Item The Role of Part-Set Cuing and Retrieval Induced Forgetting in Subjective Probability Judgments(2007-08-14) Tomlinson, Tracy; Dougherty, Michael; Psychology; Digital Repository at the University of Maryland; University of Maryland (College Park, Md.)A fundamental assumption of support theory is that unpacking an implicit disjunctive hypothesis into its component hypotheses increases its perceived likelihood compared to ratings of the implicit disjunction (Tversky & Koehler, 1994). However, recent work by Sloman et al. (2004) revealed that cuing participants with atypical exemplars from a category led to decreases in perceived likelihood. Three interpretations of this typicality effect are reviewed and three experiments are reported that examine these interpretations. Experiment 1 replicated the Sloman et al. (2004) findings but the generation data indicate that the judgment results may be due to a misinterpretation of the question. Experiment 2 adapted the retrieval-induced-forgetting paradigm and found that unpacking the implicit disjunction is affected by retrieval inducement processes, and the subjective probability judgments may be better accounted for by an averaging model. Experiment 3 indicates that these typicality effects are not observed within small judgment sets.