Additive manufacturing (AM) is the automated layer-wise fabrication of 3D objects directly from geometrical computer models. In the music technology lab, AM affords the rapid testing of enclosures and components for both sound capturing and sound producing instruments. These instruments, typically outsourced to manufacturing hubs, are now produced near the site of musical experimentation through desktop fabrication. Parametric modeling and rapid fabrication with AM accelerate the design cycle for the production of instruments for music research. AM alters the site of manufacturing, the duration between the digital sketch and its materialization, and the material constitution of the instruments we might deploy; these shifts have consequences for what gets made and in turn alter artistic practices that integrate such tool-making tools. To illustrate these affordances of AM, I describe a suite of instruments within a taxonomy of usage categories that move from reproduction of known instruments, to the augmentation of found ones and finally a phase of invention. Invention arises not simply from putting such machines in the service of novel ideas but rather from carefully examining the material outcomes of the layering process itself. The term AM is used in place of numerous alternatives to emphasize this anisotropic grain of the printed object. Although the additive method announces a powerful flexibility in the shapes it sheds, its performance in various acoustic and electroacoustic scenarios depends in part on this grain. The practice, therefore, navigates trade-offs between malleable fabrication methods for organizing material and the quality of the resulting forms for organizing sound.

Ongoing advances in DNA sequencing technologies have undoubtedly shifted genomics into the realm of big data. To cope with this data explosion and enable rapid advances in biology and medicine, we must develop scalable and efficient methods for genomic data analysis that can leverage both domain specific knowledge, as well as the latest computing platforms, such as multi-core and accelerator-based architectures and cloud computing. This thesis presents several algorithms developed with this goal in mind, focusing on key tasks performed in resequencing studies, metagenomics, and cancer genomics. First, we consider the ubiquitous and compute-intensive task of read mapping. We propose the read mapping algorithm BALAUR that can outsource a significant portion of the computation to the public cloud preserving the privacy of genomic data. BALAUR uses the MinHash technique and a coarse-grained voting scheme to map the reads, relying on the high degree of similarity between the reads and their best reference matches. BALAUR has a similar runtime to state-of-the-art read mappers in short read mapping and achieves significant speedups over existing approaches in long read mapping. We also propose the read mapping algorithm BWBBLE that maps reads to a large collection of genomes with the goal of eliminating reference bias and boosting mapping accuracy. BWBBLE creates a compressed linear representation of the collection, taking advantage of the high redundancy across the genomes. BWBBLE then uses this representation, along with an adaptation of the Burrows-Wheeler Transform search algorithm to efficiently map the reads (since all the reads can be treated independently, this is an embarrassingly parallel task). This results in significant reduction in space and runtime requirements, compared to mapping to each genome individually. Next, we consider several resource-intensive tasks in metagenomics. We introduce the framework GATTACA for fast and accurate metagenomic binning. GATTACA uses a lightweight approach for estimating co-abundance profiles across a cohort of metagenomic samples (subsequently used as features for binning), replacing read mapping with a kmer counting procedure. This approach can result in up to several orders of magnitude speedup in abundance estimation; it has also been parallelized for shared memory systems. Creating compact indices of metagenomic samples, GATTACA enables easy reuse across studies via fast downloads and small disk footprints. GATTACA also incorporates a fast procedure for metagenomic sample comparison based on MinHash that can be used for cohort selection. Finally, we focus on cancer genomics, specifically the critical task of inferring tumor heterogeneity and evolution. We introduce a novel algorithm, LICHeE, that uses variant frequencies from deep sequencing data across multiple tumor samples to infer the underlying cancer lineage tree and decompose the samples into subclonal populations. LICHeE is fast and efficiently scales with the number of input samples and variants, relying on a combinatorial formulation of the task as a search for spanning trees in a constraint network.

Biofilms are multicellular communities consisting of microorganisms enmeshed in an extracellular matrix of biopolymers. The matrix provides the community structure and cohesiveness and allows it to adhere to a variety of interfaces. Formation of a biofilm is advantageous to the microbial community, as it provides protection from external assaults (desiccation, oxidizing agents, predation), protection from host immune defenses, facilitates close cell-to-cell interactions for DNA exchange, and creates nutrient gradients that give rise to metabolic diversity within the community. These factors allow biofilms to persist in a variety of settings, ranging from large-scale industrial equipment to medical implants in the human host. In fact, many infections are now appreciated to be biofilm-related and are difficult to treat by traditional means such as antibiotics. To combat unwanted biofilms, a current strategy is to take a biophysical approach and interfere with the biofilm structure by disrupting the extracellular matrix. This strategy could revoke the survival advantages provided to the microorganisms by existing in the biofilm community. It also avoids the life-or-death pressure placed on microorganisms by traditional antibiotic treatment that gives rise to drug resistant mutations. However, to achieve this goal of targeting the extracellular matrix, we require an improved understanding of the underlying mechanical properties of the biofilm structure. In this work, we describe the use of modified rheological methods to quantify mechanical interactions relevant at all stages of the biofilm lifecycle, including: initial microbial adhesion to interfaces, maturation of the biofilm structure, and microbial dispersal. A Live Cell Monolayer Rheometer (LCMR) was used to study adhesion of uropathogenic Escherichia coli to bladder epithelial cells, the initial step in bladder infection. Quantitative mechanical measurements defined the contributions of bacterially produced type 1 pili, curli, and cellulose to bladder cell adhesion, and revealed an important role for cellulose in mediating these interactions. This novel use of live cell rheology can be expanded to study a variety of other relevant host-pathogen interactions. In a separate study, interfacial shear rheology was used to study the maturation of biofilms formed at the air-liquid interface by Vibrio cholerae, the causative agent of cholera. It was discovered that out of several known extracellular matrix components in the V. cholerae biofilm, a specific matrix protein called Bap1 contributed significantly to maintenance of biofilm elasticity, biofilm hydrophobicity, and development of a mature biofilm structure. Finally, mechanical measurements relevant to biofilm dispersal were performed using a custom-built device to apply large deformations to Bacillus subtilis biofilms formed at the air-liquid interface. These measurements revealed that biofilms exhibit non-uniform deformation due to inhomogeneous mechanical properties within the structure and can have both viscoelastic and viscoplastic characteristics. Together, these studies produced new tools in the field of biofilm mechanics and provided quantitative measurements of mechanical interactions relevant to all stages of the biofilm lifecycle.

In this work, we introduce and apply several new techniques for oil/gas reservoir optimization under uncertainty. As the first contribution, we develop a general methodology for optimal closed-loop field development (CLFD) under geological uncertainty. CLFD involves three major steps: optimizing the field development plan based on current geological knowledge, drilling new wells and collecting hard (well) data and production data, and updating multiple geological models based on all of the available data. In the optimization step, the number, type, locations and controls for new wells (and future controls for existing wells) are optimized using a hybrid Particle Swarm Optimization -- Mesh Adaptive Direct Search algorithm. The objective in the examples presented is to maximize expected (over multiple realizations) net present value (NPV) of the overall project. History matching is accomplished using an adjoint-gradient-based randomized maximum likelihood (RML) procedure. Different treatments are presented for history matching Gaussian and channelized models. Because the CLFD history matching component is fast relative to the optimization component, we generate a relatively large number of history matched models. Optimization is then performed using a representative subset of these realizations. We introduce a systematic optimization with sample validation (OSV) procedure, in which the number of realizations used for optimization is increased if a validation criterion is not satisfied. The CLFD methodology is applied to two- and three-dimensional example cases. Results show that the use of CLFD increases the NPV for the `true' (synthetic) model by 10% --70% relative to that achieved by optimizing over a large number of prior realizations. The CLFD framework includes several components, and different approaches for history matching, optimization, model selection and economic evaluation can be applied. In our second contribution, we address the problem of selecting a subset of representative geological realizations from a large set. Towards this goal, we introduce a general framework, based on clustering, for selecting a representative subset of realizations for use in simulations involving `new' sets of decision parameters. Prior to clustering, each realization is represented by a low-dimensional feature vector that contains a combination of permeability-based and flow-based quantities. Calculation of flow-based features requires the specification of a (base) flow problem and simulation over the full set of realizations. Permeability information is captured concisely through use of principal component analysis. By computing the difference between the flow response for the subset and the full set, we quantify the performance of various realization-selection methods. The impact of different weightings for flow and permeability information in the cluster-based selection procedure is assessed for a range of examples involving different types of decision parameters. These decision parameters are generated either randomly, in a manner that is consistent with the solutions proposed in global stochastic optimization procedures such as GA and PSO, or through perturbation around a base case, consistent with the solutions considered in pattern search optimization. We find that flow-based clustering is preferable for problems involving new well settings (e.g., time-varying well bottom-hole pressures) or small changes in well configuration, while both permeability-based and flow-based clustering provide similar results for (new) random multiwell configurations. We also investigate the use of efficient tracer-type simulations for obtaining flow-based features, and demonstrate that this treatment performs nearly as well as full-physics simulations for the cases considered. The various procedures are applied to select realizations for use in production optimization under uncertainty, which greatly accelerates the optimization computations. Optimization performance is shown to be consistent with the realization-selection results for cases involving new decision parameters. In the third contribution, we introduce a methodology for the joint optimization of economic project life and well controls. We present a nested formulation for this joint optimization problem where we maximize NPV, subject to the constraint that the rate of return of operations is greater than the minimum attractive rate of return (MARR) or hurdle rate. The methodology provides the optimal project life and the optimal well controls such that the maximum NPV is obtained at the end of the project life, and the rate of return of the project is essentially equal to MARR. Application of this procedure, enables avoiding situations where NPV increases slowly in time, but the benefit relative to the capital employed is extremely low. We demonstrate the successful application of this treatment for production optimization for two- and three-dimensional reservoir models.

This dissertation studies operational and policy-making problems that arise in agricultural supply chains. In the first two chapters, we explore the role of government interventions in poverty alleviation and promotion of sustainable production, respectively. In the third chapter, we investigate farmers' land allocation decisions in the face of a diversification option. Chapter 1 is motivated by the policy makers' efforts in alleviating poverty and maintaining food security by supporting poor farmers in developing countries. We investigate the effectiveness of three types of interventions, price support, cost support and yield enhancement efforts, as well as different policy implementation methods such as announcing the total budget or the unit support, in terms of their impact on farmers' incomes, consumer surplus, and return on government spending. We show that price and cost support interventions are equivalent if the total budget is public information. On the other hand, if the government announces the unit support, the benefit to different stakeholders along the agricultural supply chain depends on the market distortion created by the intervention. Specifically, in this case, price support results in greater distortion, benefiting consumers more than cost support whereas the converse holds for farmers. Furthermore, we find that under yield enhancing efforts, farmers may incur losses due to the interplay of several market and crop characteristics. Lastly, we show that interventions cannot always generate positive return from the government's perspective. Chapter 2 explores the role of policy instruments in promoting sustainable practices in agricultural production. We investigate the effectiveness of a number of policy instruments, i.e., taxes and subsidies, in terms of their impact on adoption of sustainable practices, producers' profits, consumer surplus and return on government spending. Our findings indicate that while using only taxes encourages the adoption of sustainable production, social welfare decreases as a result. Utilizing only subsidies outperforms policies that involve both taxes and subsidies in achieving higher social welfare but the converse is true in achieving a higher adoption rate. We show that zero-expenditure policies result in a decline in social welfare unless producers face financial constraints in making the costly transition to sustainable practices. Finally, we conduct a numerical study using data on conventional and organic egg production in Denmark and make policy recommendations in order to achieve the target adoption rate set by the Danish government. Chapter 3 investigates farmers' production decisions when facing different crop options. We investigate the value of crop diversification for farmers and the impact of this flexibility on the supply chain. Our results indicate that farmers' land allocation decisions depend on the total farm space (capacity) available and that diversification may not be the equilibrium outcome if the capacity is low. We find that as the profitability gap between the alternative crops increases, monoculture outweighs diversification for higher capacity values. When diversification is the equilibrium outcome, yield variability of both crops adversely affects farmers and the supply chain. On the other hand, buyers are better off when the alternative crop has high yield variability. In fact, buyers may benefit from an increase in the yield variability of the crop they buy if farmers are incentivized to limit the production of the alternative crop in order to extract the maximum revenue from the market.

The reentry blackout phenomenon affects most spacecraft entering a dense planetary atmosphere from space, due to a plasma layer that surrounds the spacecraft. This plasma layer is created by the ionization of ambient air due to shock and frictional heating created by the moving reentry vehicle, and, in some cases is further enhanced due to contamination by ablation products. The highly mobile electrons in the plasma cause a strong attenuation of incoming and outgoing electromagnetic waves, including those used for command and control, communication, and telemetry over a period referred to as the ``blackout period''. The blackout period may last up to several minutes, and at reentry speeds that may be of the order of 10 km/s, poses a serious safety hazard for the payload on board the spacecraft, especially for human spaceflight. In this work, we present a method for alleviation of reentry blackout using electric fields in a pulsed fashion. We study the reentry plasma's interaction with electronegative voltage pulses using computer simulations that incorporate models of the plasma's response to the applied electric field and interactions between the plasma sheath and the spacecraft surface. The simulations show how one can create pockets of depleted electron density in the reentry plasma sheath that may be used as ``communication windows'', thereby circumventing reentry blackout. Several parametric sweeps are also performed in order to design a blackout alleviation system. Finally, we present a discussion of experimental efforts to verify the simulation results and conclude with a conceptual design for a reentry communications blackout alleviation system based on the exclusive use of electric fields.

A very large number of practical optimization problems have been expressed as minimization of a convex objective function over a nonconvex set, specifically discrete sets such as the set of integer points. These problems arise in a variety of fields such as statistics and modeling, data analysis, and control. Two general approaches have been used for solving mixed integer optimization problems. One approach is to solve the problem globally using methods such as branch-and-bound, which suffer from exponential worst case time complexity. Another approach is to use heuristics such as semidefinite relaxation hierarchies which terminate in polynomial time, but do not guarantee finding the global solution. In this dissertation, we discuss a generic system of heuristic solutions for such problems. We describe an implementation of these methods in a package called NCVX, as an extension of CVXPY, a Python package for formulating and solving convex optimization problems. Our heuristics, which employ convex relaxations, convex restrictions, local neighbor search methods, and the alternating direction method of multipliers (ADMM), require the solution of a modest number of convex problems, and are meant to apply to general problems, without much tuning. Several numerical examples such as regressor selection, circle packing, the traveling salesman problem, and factor analysis modeling are studied. We also describe a fast optimization algorithm for approximately minimizing convex quadratic functions over the intersection of affine and separable constraints. We discuss the favorable computational aspects of our algorithm, which allow it to run quickly even on very modest computational platforms such as embedded processors. We study numerical examples including hybrid vehicle control and power converter control, and we compare our methods with existing open source and commercial alternatives. Finally we discuss how randomized linear programming heuristics can be used to solve graph isomorphism problems. We motivate our heuristics by showing guarantees under some conditions, and present numerical experiments that show effectiveness of these heuristics in the general case.

Despite a longstanding narrative celebrating the inexorable success of American colleges and universities, at certain pivotal times very powerful elites have found the higher education sector to be pitiful, broken, and in urgent need of reform. This is the story of the first iteration of that discontent. This dissertation is an account of American higher education during the period that Richard Hofstadter dubbed the "Age of Reform, " especially the years between 1890 and 1920. During these decades, a cadre of powerful and well-funded reformers clamored for sweeping changes in American colleges and universities. I name these individuals "the academic engineers." Borrowing the language and ethos of engineering, they sought to corral an unruly collection of autonomous institutions into a centralized system; to consolidate control of that system in their own hands; to create a tiered, pyramidal hierarchy of institutions; and to eliminate inefficiencies and distractions on campus. Again and again, I show how the academic engineers' efforts fell short despite their considerable wealth and power. Still, amidst their failures, we can see many key features of the modern higher education sector taking shape. Part of this dissertation's significance is its assurance that although actions always have consequences, top-down design is not destiny. The reform agenda has existed for more than a century, and it will surely continue to exist. Those of us in the rank-and-file of higher education may be best served by seeing it not as an existential threat, but as a fact of life to be assimilated, diverted, or subverted on an ongoing basis.

Cellular state is an old concept. However, scientists have only recently begun the systematic manipulation of cells to characterize and understand the functions of myriad states. As biotechnology advances enable innovative and large-scale measurements on cellular components, new biostatistical tools are required to make sense of the increased data size and complexity, which in turn augment our knowledge of cellular states. In this dissertation, I discuss my contributions to the study of cellular states from the theory and computation angles: 1) modeling and inference of regulatory gene networks with systems of nonlinear deterministic and stochastic differential equations; 2) partition-assisted clustering analysis of high-dimensional single-cell mass cytometry data; and 3) the alignment of subpopulations of cells across cytometry samples by similarity in the associated network structures. These contributions cement a platform that furthers the discussion of cellular states by framing it in both mechanistic and quantitative terms. This platform adds layers of biostatistical knowledge to Biosciences and enhances the discovery of cellular state properties.

Recent mobile vision applications demand energy-efficient real-time object detection. Specialized hardware design is needed to push the limits of both performance and energy-efficiency. While such hardware has been demonstrated for backend detection, current imager frontends consume a significant fraction of total system energy. Therefore, additional system-level energy savings may be achieved by reducing the energy requirements of frontend image capture. At the same time, it is crucial that the energy saving techniques used do not significantly degrade object-detection performance. This dissertation studies the effects of frontend imager parameters on object detection performance and energy consumption. A simulation framework, including a largescale RAW image database for object detection, is developed. And simulation results quantifying the tradeoff between pixel bitdepth and HOG-based object detection performance are presented. A custom version of HOG features based on 2-bit pixel ratios is introduced, and shown to achieve superior object detection performance for the same estimated energy compared to conventional HOG features. A frontend hardware implementation capable of extracting these features at multiple scales is proposed, and a system-level energy analysis is performed. This energy analysis suggests a potential 19X reduction in I/O energy and 3.3X reduction in backend detection energy compared to conventional object detection pipelines.