Multi-task policies enable a user to adjust their desired objective or task parameters withouthaving to train a new policy for every new desired task. In order to train multi-task policies that can generalize to unseen tasks it is common to train them on a large repository of tasks. Tasks are commonly learned with demonstrations or reward functions. However, collecting human demonstrations or instrumenting reward functions for each new task is expensive and limits scaling of multi-task policies. How tasks are specified to multi-task policies is also an important dimension that can result in expensive labor during task communication. In this thesis we explore ways to learn and specify new tasks with minimal human supervision to enable more scalable multi-task policies.

Global change has severely impacted the flow regimes of river systems across the globe, and the species inhabiting those systems now find themselves adapted to hydrographs that no longer exist. This includes many of the flow regimes in California, as streams are experiencing an altered precipitation cycle and a system of dams, culverts, and wastewater treatment plants that have transformed rivers that were once naturally flowing into occasionally concrete-lined channels. In the following chapters, I report the results of several studies aimed at understanding the relationship between flow variability and the persistence of ecological communities in river systems, particularly diatoms and benthic macroinvertebrates. In chapter one, I utilize a set of high-elevation experimental stream channels to subject the stream benthic community to a gradient of drought treatments. In this study, I found that pool habitat serves as a better refuge to benthic macroinvertebrates than riffles, contingent on the presence of hyporheic flow. Additionally, perennial flow–adapted communities appear to resist flow reduction, up to the point when surface flow is lost. My second chapter examines the relationship between benthic macroinvertebrate community composition and flow regimes, spatial connectivity, and environmental variables at large spatial scales and long temporal scales. I found that spatial processes, such as drainage density and upstream area, and flow metrics, such as flow variability and zero-flow days, best explain temporal beta diversity in macroinvertebrate communities at the multi-catchment scale. Chapter three explores the impact of urban flow disturbance, in the form of effluent discharge from wastewater treatment plants and storm runoff, on benthic diatom and macroinvertebrate communities through space and time. I found that macroinvertebrates in such systems are resistant to flow perturbations, compositionally reflected in a relatively disturbance-tolerant community, while the diatoms reflected fast-paced resilience strategies. While erratic flow shutdowns from wastewater treatment plants result in losses of taxa and lower densities of individuals, large storm events further compound flow regime disturbance in the urban stream system. Overall in this dissertation, I demonstrate that flow disturbance is a strong predictor of benthic freshwater communities, and that determining the extent of hydrological disturbance at the correct timescale is critical to the conservation of freshwater biodiversity.

DNA origami is a powerful tool for the production of self-assembled nanostructures and has shown promising applications in a wide range of fields. These structures are typically assembled through a reaction between one long single-stranded DNA (ssDNA) scaffold and many short oligonucleotide staples. In our efforts towards developing applications of DNA origami, we determined that two important properties for scaffold production include the ability to customize its size and sequence, and the ability to produce scaffolds cost effectively at scale. Several in vitro methods exist for producing custom scaffolds. However, because they rely on enzymatic processes, these methods are unreliable at large scale. On the other hand, scalable production of scaffolds based on the M13 bacteriophage genome has been demonstrated. However, the M13 genome will only accommodate insertions up to 2.5 kilobases (kb) reliably, limiting customizability. Likewise, phagemid vectors allow for the production of phage-based scaffolds with larger custom regions, but they still require 2.9 kb of fixed sequence. To overcome these issues, we developed a modified expression system using a novel phagemid, which we call pScaf. pScaf includes a standard M13 origin of replication (M13 ori) as well as a modified M13 ori which behaves as a terminator of ssDNA synthesis, allowing for packaging and export of the intervening sequence as a custom phage particle, while reducing the fixed-sequence region to only 381 bases.

Here, I discuss the steps taken to develop this novel phagemid vector. I also describe the resulting custom scaffolds which we produced using this vector and demonstrate the usefulness of these scaffolds for applications in DNA origami.

In this thesis, bidirectional isolated dc–dc converters are analyzed for MVDC applications. Medium voltage dc (MVDC) is an attractive architecture for some power systems including all-electric ship systems based on numerous advantages including increased efficiency and reduced cost. The US Navy is investing in MVDC technology for future shipboard power systems. In this dissertation, an active 5-level T-type converter is used on the primary and secondary sides of a high-frequency transformer. The modified T-type converter features higher efficiency, fault tolerant capability, and smaller filter size compared to the common dual-active bridge dc–dc converters. The operation of the proposed converter is optimized based on minimizing power losses. For the optimization, the transformer core loss, reactive power, and rms current are considered. Wide bandgap devices are used for the minimization of semiconductor switching losses. A control method based on the Fourier series and decomposition theorem is proposed. A fault-tolerant analysis is carried out for this converter and post-fault switching methods are proposed for each faulty condition. To reduce voltage overshoot, a laminated dc-link is designed, and a two-level turn-off method is used for operating frequencies around 75 kHz. The steady state and dynamic operations of the proposed structure are verified through experiments. In addition, a neural network-based fault diagnostic model is developed. By applying feature selection methods, the model achieves an accuracy of over 95%. The developed model is implemented in real-time to detect and locate open-circuit failures of the switches. The detection duration for various types of single and double failures ranges from 10 to 60 cycles.

In this work, I provide a literature review of earlier works on classification of topological phases. In doing so, we motivate the idea of topological order and phases of matter that are dependent on global geometric features of systems in consideration. I review an algebraic formalism to categorize phases of matter containing non-abelian anyons, known as Braided Tensor Categories (BTC). With the motivation to study the relationship between non-abelian anyonic and symmetry defects, I also introduce the an algebraic formalism describing the interplay of global symmetries with the anyonic degrees of freedom. This formalism is known as $G$-crossed BTC. These algebraic formalism provide observable topological quantities, such as the topological $S$-matrix, that can be probed by interference measurements. When the topological $S$-matrix of the BTC is unitary, it is said that the fusion theory is said to be modular. Hence, one calls the algebraic formalism a Modular Tensor Category (MTC). I present the results of our work showing that one can construct projective representations of the mapping class group of the punctured torus from the topological formalism given by $G$-crossed BTC. This is pivotal in constructing a $G$-crossed BTC for surfaces of higher genus, as one can build surfaces of arbitrary genus by gluing many punctured tori. Throughout my mathematical analysis, I also show that modularity of BTCs are not destroyed by the presence of symmetry defects.

In this thesis, we explore topological phases, study their properties, and present some novel extensions of them. Our study of topological phases begins on simplified theoretical models and provides a tractable setting to discuss the otherwise abstract mathematics involved, while maintaining the power to capture many experimentally salient features of these phases. We generalize these theoretical models and discover new phases of matter, topological flux phases, which are topological phases with a uniform anyonic flux. Other extensions of topological phases we study are fractionalized Fermi liquids, which are gapless topological phases with non-trivial interactions between gapless and topological sectors. Finally, we also focus on one peculiar property of topological phases, topologi- cal entanglement entropy, which captures the fact that some information is distributed globally in topological phase and can only be accessed with a topologically non-trivial measurement.

Quantum computing is a revolutionizing technology on many fronts. Topological quantum computing is an especially attractive platform for this due to its innate robustness to local sources of noise and scalability. As the advent of quantum computing technologies nears, it becomes increasingly important to both optimize as much as possible the basic operations and to characterize them. In this thesis we study measurement-only topological quantum computing with Majorana zero mode (MZM) based qubits, particularly on their incarnation as the boudnary defects of $1d$ topological superconducting nanowires. The specific qubit device we consider are comprised of collections of such wires connected together on a single island such that 4 or 6 MZMs are hosted (tetron and hexon qubits respectively).

This thesis focuses on optimizing and characterizing the basic operations needed for quantum computing with such MZM qubits, concentrating on hexons which are the smallest MZM qubit allowing measurement-only gates on a single island. We begin by optimizing the Clifford gates for a single hexon qubit and for the controlled-Pauli, $\mathsf{W}$, and $\mathsf{SWAP}$ gates on two hexon qubits. A brute-force searched is performed with respect to a physically motivated example cost function. Tools and techniques are developed to aid in the compiling and optimization of measurement sequences. Utilizing the developed tools, we apply it towards the optimization of the stabilizer measurements for the surface code, an especially useful quantum error correcting code for the present MZM setting. Finally, we discuss adapting the technique of randomized benchmarking to the hexon setting, addressing simplifications and extensions that a hexon qubit can offer.