UC Santa Cruz

In California both min’yō and Tsugaru shamisen practitioners occupy two disparate diasporic intercultural spaces for Japanese folk music. Based on fieldwork conducted in California music and dance affinity groups between 2018 – 2023, this study of Japanese folk music and dance argues these two practices and communities demonstrate different types of musical resiliency during precarious times while they engage with the differing desires of institutions, participants, and audiences in expressing a Japanese folk musical past. In precarity, the min’yō and Tsugaru shamisen groups demonstrate resiliency in adapting to shifting conceptions of identity whilst experimenting with alternatives to the current promotion practices of state institutions and NGOs. The min’yō and Tsugaru shamisen communities work to sustain the spatial aesthetics of Japanese folk music and dance which are tied to specific places and practices in Japan. While cultivating this aesthetic, practitioners move across different transmission spaces with masters and peers both in-person and digitally distanced. The contemporary place of min’yō and Tsugaru shamisen communities in California is informed by a Japanese American history of displacement. These groups exhibit resiliency in negotiating the transitions and transformations of Japanese folk music and dance.

Finite escape means the occurrence of an infinite value in the solution of a time-varyingdifferential equation. Finite escape can occur in the computation of either the Kalman or the Kalman-Bucy filter because the gain is time-varying. When no escape occurs it is analogous to the Heisenberg uncertainty principle [1] in atomic physics. Three noisy examples are given: a single integrator, a double integrator, and a linear oscillator. Finite escape cannot happen in the single integrator or the underdamped linear oscillator, but can happen in the double integrator and undamped linear oscillator. Therefore, finite escape can occur in the estimation of any noisy dynamic system. Except in special situations, it is impossible to achieve certainty in the determination of all state variables using neural nets, machine learning, or artificial intelligence even with an infinite amount of data. Conditions for finite escape to occur are given. Finally, practical solutions for escape are considered for the linear oscillator.

Filarial nematodes are human parasites that infect millions of people across the globe and cause debilitating diseases such as Elephantiasis and African River Blindness. These worms share a symbiotic relationship with Wolbachia, an obligate, alpha-proteobacteria endosymbiont, and rely on these bacteria for survival and proper embryogenesis. Taking advantage of this crucial symbiosis, efforts to identify drugs that kill the adult worm by targeting Wolbachia have proven to be promising. Here, I describe the discovery and optimization of quinazolines CBR417 and CBR490 that, with a single dose, achieve >99% elimination of Wolbachia in the in vivo Litomosoides sigmodontis filarial infection model. These potent quinazolines were identified by pairing a primary cell-based high-throughput imaging screen with a secondary ex vivo validation assay to rapidly quantify Wolbachia elimination in Brugia pahangi filarial ovaries. To better understand the relationship between Wolbachia and its worm host, adult Brugia pahangi were exposed to varying concentrations of common antibiotics in vitro and assessed for Wolbachia numbers in the germline tissue. Surprisingly, we found that worms treated with higher concentrations of antibiotics had higher Wolbachia titers, and antibiotics given at low concentrations reduced Wolbachia titers. This counterintuitive dose response is known as the “Eagle effect” and the presence of this effect in Wolbachia suggests a common underlying mechanism that allows diverse bacterial and fungal species to persist despite exposure to high concentrations of antimicrobial compounds. To our knowledge this is the first report of this phenomenon occurring in an intracellular endosymbiont, Wolbachia, in its filarial host. While several studies have shown that novel and FDA-approved antibiotics are efficacious at depleting the filarial nematodes of their endosymbiont, thus reducing female fecundity, it remains unclear if antibiotics can permanently deplete Wolbachia and cause sterility for the lifespan of the adult worms. We investigated the long-term effects of the antibiotic, rifampicin, in the Brugia pahangi jird model of infection. Initially, rifampicin treatment depleted Wolbachia in adult worms and simultaneously impaired female worm fecundity. However, during an 8-month washout period, Wolbachia titers rebounded and embryogenesis returned to normal. Clusters of densely packed Wolbachia within the worm’s ovarian tissues were observed by confocal microscopy. The number, size, and Wolbachia density of these clusters were not diminished despite large doses of rifampicin antibiotic. This finding suggests that these clusters may serve as privileged sites that allow Wolbachia to persist in worms while treated with antibiotic. Lastly, I define the cellular characteristics of these clusters, which fit the definition of endosymbiont bacteriocytes, and I identify drugs that target them. Nascent bacteriocytes arise in newly formed sheath cells adjacent to the distal tip cell of the Brugia pahangi germline. They dramatically enlarge but do not appear to disrupt the integrity of the sheath cells. We determined that the Wolbachia within bacteriocytes are either in a quiescent form or replicating at a very low rate. These Wolbachia-based bacteriocytes are present in Brugia malayi, one of the nematode species which cause human filariasis, as well as B. pahangi. Screens of known antibiotics and other drugs revealed two drugs, Fexinidazole and Corallopyronin A, have strong anti-bacteriocyte efficacy.

Coastal wetlands are highly productive ecosystems and can store large amounts of carbon (C). However, decomposition processes in coastal wetlands also produce and emit greenhouse gasses (GHG), such as methane (CH4) - a potent greenhouse gas that could offset C storage in the wetland soil. Often a patchwork of vegetation and open water, coastal wetlands exhibit strong biogeochemical heterogeneity, resulting in elevated CH4 flux (FCH4) at certain times and locations. These points of elevated FCH4, termed “hot spots and hot moments" (HSHM), experience biogeochemical rates so high they can disproportionally contribute to annual flux rates. Despite the broad utilization of the term HSHM, there is no standardized, statistically rigorous method for identifying HSHM and quantifying their impact on ecosystem processes. Furthermore, the conditions that trigger HSHM of FCH4 are poorly understood, and hot moments are often excluded from wetland FCH4 upscaling and predictive modeling. This study presents a comparative analysis of standard HM identification techniques to find the best HM detection method for coastal wetlands and formalize HM identification best practices. We found that using a rolling Z-score threshold to identify hot moments from eddy covariance (EC) flux data was most suitable for coastal wetlands. Using this approach, we flagged hot moments at nine wetlands in the San Francisco Bay-San Joaquin River Delta (Bay-Delta). We then used the identified HMs to train several data-driven Random Forest (RF) models that leverage EC data to predict the occurrence of HMs. The best performing RF accurately (79%) captured HM absence/presence in the Bay-Delta region, and the relative importance of predictive environment parameters in the model shed light on the best predictors for HM. The method comparison in this study provides a best practices workflow for researchers when defining HSHM, and the RF HM model provides an upscaling methodology that could be used to predict the occurrence of HM FCH4 at sites without EC towers. Thus, the HM identification methodology and the predictive model present a valuable tool for wetland managers and restoration planners who can use the information to prioritize time and resources for mitigating and preventing these rare but high-impact emission events.

This dissertation explores the utilization of internet-connected devices to addressthe accessibility challenges associated with microscopy in the field of biology. Tra- ditional microscopy devices are often costly, limiting their widespread adoption in laboratories. To overcome this barrier, we developed a cost-effective multi-well imaging device integrated with an open-source pump and stimulation system. This device facilitates longitudinal cell studies and reduces the time required for capturing cell images, thereby enhancing efficiency for biologists. Moreover, our internet-connected imaging system serves as an innovative ed- ucational tool for aspiring scientists. Through remote project-based learning ac- tivities and engaging serious games, complex biological concepts become more accessible, fostering enthusiasm and understanding among high school and under- graduate students. The dissertation discusses the application of this system in an educational setting and the implementation of an algorithm for determining organoid size through computer vision. Additionally, it introduces an interactive virtual lab, designed to provide a low-stakes environment for students to learn scientific and laboratory safety protocols. Overall, this work highlights the potential of internet-connected devices in revolutionizing both biological research and education. By making microscopy more accessible and engaging, it empowers scientists and students alike to explore the intricacies of biology with enthusiasm and curiosity.

In the modern Lambda-CDM model of cosmology, galaxies form in the centers of overdense regions in the cosmic web, known as dark matter halos. The formation and evolution of galaxies are believed to be connected to the formation and evolution of the halos they occupy. This concept is referred to as the galaxy--halo connection, and it provides us with an avenue for understanding the complex physics involved in galaxy formation. Because we assume every galaxy is located in the center of a halo, drawing parallels between the spatial distributions of galaxies and halos is an effective way of illuminating how halo properties may be connected to galaxy properties. However, three-dimensional spatial information is difficult to obtain accurately in the real Universe, as all information must be extracted from the emitted light of distant galaxies. In this paper, we apply the stochastic order redshift technique (SORT) to mock redshift surveys to test how well it recovers the true distribution of galaxies. SORT relies on a small (10%) reference sample of high-quality redshifts that outline the underlying structure of galaxies to determine new estimates of low-quality redshifts. We find that SORT overall improves redshifts, recovers the redshift-space clustering on scales > 2.5 Mpc/h, and provides improved estimates of local densities. Then, we study the clustering properties of central SDSS galaxies as a function of specific star formation rate (sSFR). We find that central galaxy auto-correlations show little dependence on sSFR, with the established result of quiescent galaxies clustering more strongly than star-forming galaxies attributable to satellites. Because halo assembly history is known to affect distinct halo clustering, this result implies there is little net correlation between halo assembly history and central galaxy sSFR. We also find that cross-correlations of centrals with satellites increase with lower sSFR, suggesting that quiescent centrals have more satellites than star-forming centrals of the same mass. We compare our findings to the predictions of empirical models of sSFR using the Bolshoi-Planck N-body simulation and find that models dependent on halo assembly history disagree with observations while a model independent of halo assembly history reproduces well the observed clustering properties of centrals.

Reachability analysis is a method to guarantee the performance of safety-critical applications such as automated driving and robotics against dynamic uncertainties. The main object of study is the reach set, defined as the set of states that a controlled dynamical system may reach at a future time, depending on a set-valued evolution of uncertainties. We develop the theory and algorithms for learning the reach sets of full state feedback linearizable systems---an important class of nonlinear control systems, common in vehicular applications such as automobiles and drones. These reach sets, in very general settings, are compact but nonconvex. The new idea we propose is to compute these reach sets in the associated Brunovsky normal coordinates, and then transform the sets back to the original coordinates via known diffeomorphisms. Our algorithms exploit learning-theoretic ideas to provide probabilistic guarantees on the computed sets.

As a by-product of our analysis, we uncover the exact geometry of the integrator reach set with compact set-valued inputs. These exact results include the closed-form parametric and implicit formulae for the boundaries, volumes, and widths of the integrator reach sets. The exact parametric formula for the boundary admits an integral representation involving the boundary of the compact input sets. The exact implicit formula is given by the vanishing of certain Hankel determinants. These results on integrators should be of independent interest, serving as benchmarks for quantifying the conservatism in reach set computation algorithms.

Our geometric analysis also helps clarify a taxonomy, i.e., what kind of compact convex sets can the integrator reach sets be. We show that the integrator reach sets resulting from arbitrary, time-invariant, compact input sets are zonoids and semialgebraic, but not spectrahedra. The integrator reach sets resulting from arbitrary, time-varying, compact input sets are shown to be zonoids but not semialgebraic in general.

We detail how these geometric results enable the semi-analytical computation of the reach set of any controllable linear time invariant system, as well as the reach sets of full state feedback linearizable systems.

Leveraging an Isomorphism between compact sets and their support functions, we also propose a data-driven method for learning any general compact set.

This is useful for learning compact sets such as reach sets, maximal control invariant sets, region-of-attraction that are related to an underlying nonlinear dynamical system but an analytic model for the dynamical nonlinearities are unavailable. Our results show that the proposed geometric learning ideas can be efficient when we only have access to simulated or experimentally observed data. We demonstrate computational learning of these compact sets by carrying out regression analyses on their support functions using finite data sets.Finally, we outline the directions for future research.

Current smartphone-based localization systems, primarily designed towards sighted individuals, offer wayfinding services by tracking a user's path. However, this design overlooks the unique navigation needs of blind individuals who utilize long canes or guide dogs and have distinct movement patterns. To bridge this gap, this thesis introduces novel localization techniques tailored for blind pedestrians in both indoor and outdoor settings. These techniques avoid the need for BLE beacons and Wi-Fi, as well as camera-based systems, all of which are impractical for blind users. Instead of these options, the proposed methods utilize IMU sensors, allowing users to conveniently place their phones in their pockets without the requirement of any external infrastructure.

Indoor localization in the absence of maps is addressed in this thesis through a unique combination of a Mixture Kalman filter and a GRU-based straight walking detector. Together, these form a two-stage turn detector that operates under the assumption that corridor intersections occur at 45° or 90° angles. In situations where maps are accessible, the research incorporates two Pedestrian Dead Reckoning (PDR) methods with the map data via a particle filter. In outdoor settings, this thesis expands the use of IMU sensor data by integrating it with GPS signals through a particle filter. This method creates a flexible model effective in both open areas and in environments with wall constraints, as specified by maps. Comprehensive testing of these systems involved trials with the WeAllWalk dataset, containing data from visually impaired walkers, and user studies conducted using two separate iPhone applications for indoor and outdoor localization. Results from these tests clearly demonstrate the effectiveness of the proposed localization solutions.

In the context of artificial intelligence (AI) or machine learning (ML), we speak of the "alignment" of an AI system's behavior with human goals, values, and ethical principles. "The alignment problem" has proven challenging, and as the capabilities and applications of AI rapidly advance, the shortcomings of standard solutions are increasingly consequential. This dissertation focuses on an often overlooked but critically important complication to the alignment problem: Socially-consequential AI systems affect their environment (involving, for example, human populations) and are therefore subject to dynamical feedback driven by other agents. We address three central questions: (1) As intelligent agents adapt to each other, does a system aligned using current leading approaches remain aligned? (2) Can we anticipate and utilize adaptive agents' reactions to data-driven policy to achieve aligned objectives dynamically? (3) How can we guarantee alignment for AI systems that interact with complex, multiagent environments that are difficult to model or predict? We will address these questions using the theoretical framework and experimental tools of machine learning---integrating concepts from dynamical systems, evolutionary game theory, constrained optimization, and control theory. We hope to demonstrate that a dynamical systems approach to deployed AI is not only necessary but beneficial to the goal of alignment.

Systems of hyperbolic conservation laws (HCLs) commonly arise as mathematical descriptors of the natural world, and are particularly ubiquitous in fluid dynamics. These laws appear as complicated and highly nonlinear partial differential equations describing the evolution of fundamental conserved quantities such as mass, momentum, and energy. Solving these equations analytically is entirely intractable for all but the simplest cases, and investigating problems with real world importance falls to numerical approaches more and more frequently. Most HCLs exhibit rich dynamics with complicated smooth flows and discontinuities coexisting, often with shocks arising frominitially smooth data. Designing numerical schemes that can efficiently and accurately represent smooth phenomena, while also remaining robust and reliable in the vicinity of shocks, is very challenging. Finite volume methods are one particularly useful approach to designing such methods as conservation is enforced discretely, and discontinuities can be represented quite naturally. An unfortunate drawback of these methods is that achieving high-order accuracy in multiple space dimensions is difficult. This dissertation overcomes these challenges by developing a kernel-based non-polynomial reconstruction scheme that is manifestly multidimensional. This scheme is first posed as a linear recovery problem in a reproducing kernel Hilbert space. This linear reconstruction method is then cast into a weighted essentially non-oscillatory (WENO) framework so that it may represent both smooth and discontinuous data. This scheme is then incorporated into solvers for the compressible Euler equations, compressible Navier-Stokes equations, and ideal magnetohydrodynamics (MHD) equations. In doing so, a novel set of variables that are more suited to multidimensional reconstruction, dubbed the linearized primitive variables, are introduced. Troubled cell indicators are developed that allow for a more accurate and efficient treatment of smooth solutions in an entirely automatic fashion. Positivity preserving limiters are also incorporated, and allow for the evolution of flows with extremely strong shocks. A highly parallel multi-GPU implementation is provided, and the proposed method is tested against a variety of stringent benchmark problems.