Skip to main content
Open Access Publications from the University of California

UCLA Electronic Theses and Dissertations

Sensitivity Analysis and Uncertainty Quantification in Reduced-Order Monopropellant Catalyst Bed Model


The present study replicates a 1D, steady, chemically-reacting, reduced-order model for hydrogen peroxide flow through a monopropellant catalyst bed as described in Pasini et al. [9] with model validation completed by comparison with both model data and experimental data from Jung et al. [10]. Adaptations were made to improve heat transfer capabilities within the model and to adapt the model such that a hydroxylammonium nitrate and water mixture could be used as the propellant. Polynomial chaos expansion was implemented to decrease sampling time in order to perform non-deterministic analyses including: quantification of global sensitivities using Sobol indices, construction of axial property profiles with uncertainty envelopes for random physical inputs, and construction of posterior probability distributions with confidence intervals for variation in chemical tuning parameters. Results from the study show that model behavior is primarily governed by propellant mass fraction and activation energy of the global reaction. Additionally, posterior distributions indicate that activation energy and number of active sites per volume are related by a logarithmic family of solutions as a result of the reaction advancement gradient form in the model.

Cover page of Effective and Efficient Representation Learning for Graph Structures

Effective and Efficient Representation Learning for Graph Structures


Graph structures are a powerful abstraction of many real-world data, such as human interactions and information networks. Despite the powerful abstraction, graphs are challenging to model due to the high-dimensional, irregular and heterogeneous characteristics of many real-world graph data. An essential problem arose is how to effectively and efficiently learn the representation for objects in graphs. In this thesis, both the effectiveness as well as efficiency aspects of the graph representation learning problem are addressed. Specifically, we start by proposing an effective approach for learning heterogeneous graph embedding in an unsupervised setting. Then this is generalized to semi-supervised scenario where label guidance is leveraged. The effective graph representation learning models are followed by efficient techniques, where we propose efficient sampling strategies to improve the training efficiency for content-rich graph embedding models. Finally, to reduce storage and memory cost of the embedding table used in various models, we introduce a framework based on KD code, which can compress the embedding table in an end-to-end fashion. We conduct extensive experiments on various real-world tasks on graph data (e.g. anomaly detection, recommendation and text classifications), and the empirical results validate both effectiveness as well as efficiency of our proposed algorithms.

The Effect of Transportation-Related Air Pollution on the HPA Axis and Adverse Birth Outcomes


Exposure to air pollution causes dysfunction in human physiology, including in the endocrine system and reproductive process. Transportation-related air pollutants are products of combustion engines, which can come from mobile ground and air sources. Los Angeles County is home to the world’s fourth busiest passenger airport and almost 22,000 miles of maintained roadway, of which approximately 500 miles are freeway. Communities located proximal to these sources experience a considerable health burden from these exposures, especially among sensitive populations like children and pregnant mothers.

The first study of this dissertation examines the role of traffic-related air pollution in dysregulation of the hypothalamus-adrenal-pituitary gland (HPA) axis. Subjects were adolescents enrolled in the Los Angeles Family and Neighborhood Survey (LAFANS) between 2006 and 2008. We built a land use regression (LUR) model to estimate chronic nitrogen dioxide (NO2) exposure, a marker of traffic-related air pollution. We then generated model-based estimates of the association between the cortisol, a measure of HPA axis functioning, and NO2 exposure one year prior to cortisol sampling. Our results indicate that increased exposure to NO2 was associated with a flattened diurnal slope of cortisol, an indicator of an abnormal cortisol response. We hypothesize that this may be a mechanism through which air pollution may affect respiratory function and asthma in adolescents.

The second and third studies of this dissertation use birth outcome data from birth certificates supplied by the California Department of Public Health and exposure data from novel ultrafine particle (UFP) dispersion model. The second study estimates the association between in utero exposure to aircraft-related UFPs from planes landing at Los Angeles International Airport and preterm birth (PTB). Women were included in this study if they delivered babies between 2008 and 2016 and lived within 15 km of the airport. Controlling other pollutants and demographic factors, we found that the risk of PTB was elevated among women more highly exposed to aircraft-related UFPs during their pregnancy. This association was strongest among foreign-born women, especially those of Hispanic and Asian descent.

The third study is a causal mediation analysis, that examines the role of pregnancy-induced hypertension (PIH), including preeclampsia, as a mediator in the previously described relationship between UFP and PTB. We used a weighted marginal structural models approach to estimate the natural direct and indirect effects of UFP on PTB risk, with PIH as a mediating factor. We found that approximately 13% of the UFP-PTB relationship was mediated through UFP exposure-caused hypertensive disorders. We hypothesize that UFP exposure may lead to an inflammatory response which puts expectant mothers at greater risk for hypertensive disorders that can lead to iatrogenic preterm births.

In conclusion, our findings present evidence of the detrimental effects of exposure to air pollution both in utero and during adolescence. These environmental exposures are widespread in urban settings like Los Angeles and impact large populations. The burden of these health effects is often carried by low socioeconomic status communities co-exposed to other pollutants. This research exposes potential processes through which air pollution impacts human health, adding plausible biologic mechanisms that strengthen the understanding of how transportation-related emissions impact a city’s most vulnerable groups.

Cover page of Study of Polyion Complex Structure Formation from Mixing Oppositely-Charged Block Copolypeptides

Study of Polyion Complex Structure Formation from Mixing Oppositely-Charged Block Copolypeptides


Synthetic polypeptides are a versatile class of biomaterials with many interesting properties such as biodegradability, biocompatibility and ordered secondary conformations. In particular, block copolypeptides with well-defined block composition and versatile selection of amino acid constituents allow for controlled assembly into supramolecular structures such as micelles, vesicles and hydrogels. In recent years, polyion complexation has been developed as a new strategy for supramolecular structure assembly, resulting in formation of unique polyion complex (PIC) systems that have seen growing applications in drug delivery and gene therapy. However, the usage of PIC assembly in controlling block copolypeptide supramolecular structure formation has been largely unexplored. This dissertation will focus on the study of polyion complex (PIC) structure formation by mixing oppositely charged block copolypeptides.

Synthetic diblock copolypeptides were developed to incorporate oppositely charged ionic segments that form β-sheet structured hydrogel (DCHPIC) assemblies via polyion complexation when mixed in aqueous media. The polyionic block length as well as polymer concentration can be used to tune hydrogel properties. The PIC hydrogel system has self-healing properties, microporous architecture, and stability against dilution in aqueous media. Neural stem progenitor cells were also successfully loaded into the hydrogel with good cell viability. Together, these promising attributes and unique features of the β-sheet structured PIC hydrogels highlighted their potential applications as carriers for stem cell therapy.

Diblock (DB), triblock (TB) and pentablock (PB) copolypeptide PIC hydrogels with identical overall amino acid compositions and ionic block lengths were assembled and their mechanical properties were compared. Specifically, the pentablock copolypeptides were designed to be equivalent to two connected triblock copolypeptides. As a result, PB hydrogels have demonstrated drastic improvement of mechanical properties over the DB and TB hydrogels. Furthermore, low concentrations of cationic PB components can be incorporated within the DB or TB hydrogels and act as linkers to significantly increase mechanical properties.

A dual network physically cross-linked hydrogel (DCHDN) was developed that consists of two separate interpenetrating diblock copolypeptide networks based on discrete modes of assembly: polyion complexation (DCHPIC) and hydrophobic association (DCHMO). The PIC precursors were mixed within a preformed amphiphilic hydrogel to give hydrogels with two distinct networks. The DCHDN components were shown to have synergistic effects that significantly enhanced mechanical properties of the overall system. The PIC component imparts its stability against dilution to the DN hydrogel system while the amphiphilic component introduces hydrophobic domains within the network that potentially allow for hydrophobic cargo encapsulation. Contrary to many reported dual network hydrogels systems, DCHDN retains the self-healing properties of its components, which makes this hydrogel system a potential injectable carrier for controlled release applications.

PIC diblock copolypeptides have been synthesized, assembled and characterized to form assemblies. Assembly size and structure can be tuned by varying the poly(ionic) block lengths and chirality. PIC assemblies were found to have core-shell micellar structures by electron microscopy and confocal imaging. Potential use of these assemblies for protein delivery was explored with lysozyme as the model protein. The polypeptide-protein complex formed assemblies that are stable under physiological salt and osmotic conditions.

Cover page of Brain-Mimetic Hydrogel Platform for Investigation of Glioblastoma Drug Resistance

Brain-Mimetic Hydrogel Platform for Investigation of Glioblastoma Drug Resistance


Glioblastoma (GBM) is the most lethal and malignant cancer originating from the central nervous system. Even with intense treatment involving surgery and radio-chemotherapy, median survival after prognosis remains within 12 months, as GBM constantly develops resistance to common therapies. Many novel therapies developed for GBM have shown promising results in in-vitro studies, but unfortunately failed in actual clinical practices, partially because traditional model systems failed to recapitulate the microenvironment surrounding GBM tumors. Therefore, we posit that unique brain extracellular matrix (ECM) facilitates therapeutic resistance in GBM. To study this problem, we investigated ECM deposition in GBM patient samples and fabricated brain-mimetic, orthogonally tunable hydrogel system in which to culture patient-derived GBM cells in 3-dimensional manner. To validate our novel ex-vivo culture system, genomic sequencing and gene expression profiling were performed for comparison with traditional in-vitro culture and animal xenograft models. At the same time, cell viability, proliferation and markers for cancer stem cell were assessed. Our model system was used to study the therapeutic response of GBM cells to commonly used therapeutics and to investigate resistance mechanisms. We found GBM cells displayed drug response kinetics comparable to in-vivo xenograft models. We also found novel molecular mechanisms describing how unique brain matrix facilitates therapeutic resistance through corresponding receptors in our 3D culture models. By utilizing novel engineered platforms to study drug resistance, we are able to uncover mechanisms that could not be observed through traditional methods.

Cover page of Geologic Mapping and Geophysical Modeling of the Surface of Ceres: Insights into the Structural, Mechanical, and Compositional Properties of the Solar System’s Innermost Dwarf Planet

Geologic Mapping and Geophysical Modeling of the Surface of Ceres: Insights into the Structural, Mechanical, and Compositional Properties of the Solar System’s Innermost Dwarf Planet


When NASA's Dawn mission arrived at Ceres on March 6, 2015 it made history by becoming the first spacecraft to enter orbit around a second extraterrestrial object after leaving the asteroid Vesta in September 2012. Dawn thoroughly investigated the surface and deep interior of the dwarf planet Ceres, the largest object in the main asteroid belt, through a series of successively lower mapping orbits until its end of mission on November 1, 2018. Prior to Dawn’s arrival Ceres was known to be the largest C-type asteroid, and was suspected of being rich in water ice and other hydrated materials. As a putative remnant of the earliest phases of rocky planet formation, Ceres was thought to contain clues as to how planetesimals accreted and how volatiles arranged themselves throughout the inner solar system during the tumultuous era of planet formation. The Dawn mission’s goals were to further elucidate the structure and composition of the early solar system, which would lead to an increased understanding of the conditions present during terrestrial planet formation, and to determine the chemical, geological, and structural nature of the largest surviving planetary embryos: Vesta and Ceres. At Ceres, this was accomplished by meticulously characterizing the surface geology, surface and near-surface geochemistry, and interior structure via a combination of photo geology; visible, infrared, and nuclear spectroscopy; and gravimetry. This dissertation contributes to the objectives of the Dawn mission by aiding in the global geologic mapping effort of Ceres, identifying and classifying geological features indicative of its compositional and physical properties, and then applying geophysical techniques to these features in order to estimate these properties, particularly the water ice content of the near-surface. Understanding the quantity and distribution of water ice in the upper layer of Ceres is paramount for understanding both its geochemical evolution and the nature of the early solar system. The investigations presented in this dissertation reveal that Ceres is ubiquitously covered in geologic features suggestive of significant quantities of near-surface ground ice, namely large constructional mountains and hills, and a broad spectrum of lobate flow and mass wasting deposits. The observed mass wasting features exhibits physical characteristics and runout efficiencies similar to many ground ice mediated flows on Earth, Mars, and Iapetus such as long run-out landslides and frozen debris flows. Additionally, many craters on Ceres were observed to emanate fluidized appearing ejecta similar to examples found on Mars, Ganymede, and other icy worlds. Analyzing the mobilities of these ejecta using a kinematic-dynamic sliding ejecta emplacement model indicated that the cerean crust is significantly weaker than competent silicate rock at impacting conditions, and that the frictional properties of its surface are consistent with a rock-ice mixture. Finally, a unique fractured terrain named Nar Sulcus was identified on Ceres’ southern hemisphere that displayed topography suggestive of elastic flexure. By applying a flexural-cantilever model to the observed topography, the flexural rigidity of the cerean crust was estimated to be similar to those of outer solar system moons such as Europa, Ganymede, and Enceladus, which are several orders of magnitude less rigid than the terrestrial planets. From the aforementioned observations and investigations, the ground ice content of the cerean crust is estimated to be ~30-70 vol %, although there is likely significant regional heterogeneity in its distribution. This result is significant as it independently supports the interpretation that Ceres is a water rich dwarf planet, and that large quantities of ice can be sequestered within massive C-type asteroids over geologically long time periods. This is particularly exciting as carbonaceous chondrites are the most spectrally similar meteorites to C-type asteroids, and their water is the most isotopically similar to the Earth’s oceans. In light of the results presented in this dissertation, and by a myriad of other authors, it increasingly appears that a significant portion of the Earth’s water was likely delivered by Ceres-like asteroids.

Cover page of Helminth infection and treatment among pregnant women in the Democratic Republic of Congo: An examination of associated risk factors, co-morbidities, and birth outcomes

Helminth infection and treatment among pregnant women in the Democratic Republic of Congo: An examination of associated risk factors, co-morbidities, and birth outcomes


Helminth infections have an extremely high burden of disease, infecting over a billion people globally. Yet helminthiases remain heavily neglected in research and intervention efforts, especially in women of childbearing age. The exclusion of pregnant and breastfeeding women from preventive chemotherapy campaigns has exacerbated the problem in places like the Democratic Republic of Congo (DRC), a resource-poor, helminth endemic country with a high national fertility rate. The overarching aim of this dissertation is to better describe the landscape of helminthiases in DRC’s pregnant population, elucidating effects of prenatal infection on both maternal and neonatal health. Chapter 1 provides a summary of disease pathogenesis, prevention and control strategies for schistosomiasis and soil-transmitted helminths. Chapter 2 describes the prevalence, risk factors, and symptoms associated with urogenital schistosomiasis in a cross-sectional survey of women attending antenatal clinics in southeastern DRC; poor symptom recognition and a three-fold increase in the odds of sexually transmitted co-infections were identified amongst mothers harboring S. haematobium. Chapter 3 utilizes causal inference methods to examine the longitudinal effects of prenatal schistosomiasis on downstream offspring health, finding no distinction in the risk of adverse birth outcomes between mothers with treated infection and uninfected controls. Chapter 4 explores predictors and birth effects of prenatal anthelminthic use at scale, finding that in a nationally representative survey of mother-child pairs in DRC, indiscriminate anthelminthic drug use is unevenly distributed across sociodemographic lines and associated with significantly reduced odds of neonatal death. The findings of this dissertation reiterate the vulnerability of mothers and their offspring to unmitigated helminthiases, as well as the neutral or beneficial effects imparted by prenatal deworming. Given the evidence amassed herein, expansion and institutionalization of preventive chemotherapy at the national level—including the systematic incorporation of pregnant women in mass drug administration campaigns throughout DRC— is warranted.

Cover page of Salon Safety: Community-Engaged Approaches to Workplace Safety Interventions

Salon Safety: Community-Engaged Approaches to Workplace Safety Interventions


In California, the salon industry represents a significant small-business sector. Working in

these salons are cosmetologists who are exposed to a wide array of occupational hazards at

work. Toxic chemicals, musculoskeletal disorders, and psychological demands in the workplace are just a few of the hazards experienced by beauty care workers. The beauty products marketed to and used by Black women have been found to contain potentially harmful ingredients.

Black hair-salon workers face serious health hazards from these products they use

on clients and other health hazards at their work. Knowledge on this issue, as it relates to

Black hair care professionals and potential intervention methods, is extremely limited. This dissertation includes three studies that sought to understand the occupational health status

of Black salon workers in the Los Angeles region, identify workplace intervention strategies

tailored to small businesses and pilot a community-engaged intervention program aimed at

reducing workplace injuries and illnesses in the salon.

Based on the first study, a lack of proper health and safety training and personal protective

equipment use within the salon worker community was found. Additionally, it was

found that there was a willingness by stylists to learn more about workplace hazards and

how to mitigate their risks. The conclusion of this study demonstrated a need for additional

community-based studies with Black salon workers on workplace health intervention methods.

In the second study, it was found that the process of developing and facilitating an intervention program for small businesses required an understanding of the community being

served, developing a relationship with the community, building partnerships, and addressing

barriers to information. From this second study the use of community partnerships and

intermediates in the promotion of safety and environmental practices was highlighted as

instrumental for success.

In the third study, it was found that a community-engaged approach in the development

of a personal protective equipment use intervention program led to favorable results including

an increase of salon safety knowledge and personal protective equipment use among Black


Taken together, these research studies provide clear insights into comprehensive approaches

for targeted occupational safety intervention programs aimed at underserved worker


Cover page of Fundamentals, Speculation, and Seasonal Correlation in Commodity Markets

Fundamentals, Speculation, and Seasonal Correlation in Commodity Markets


The understanding of agricultural commodity financial markets has become of significant interest, given the increasing attention both the private and public sector have been giving to them. Financial investment in these goods has increased exponentially over the past 15 years. Food prices have reached significantly high levels. Therefore, my dissertation focuses on what is a relevant not only for the literature but also for both public and private institutions. I start by working within a competitive storage framework, as is usual for the literature. I then make assumptions and change the timing and information structure to match realistic aspects, therefore obtaining partially different theoretical results. I then intend to test these results by contrasting them with publicly available data regarding prices, production, consumption and information. Since the model is designed to match real-life aspects of the market, it can be applied to identify and measure fraction of price changes due to each different fundamental. For example, one recurring and important aspect of the data is that, in general, storage models predict excessively stable prices. That is, standard deviations are higher in the data than those compared to simulations in the models. Volatility is important since it can have potential welfare effects on both consumers and producers. Therefore, it is an issue that deserves attention. Moreover, related to this, in the past 15 years financial investment in these markets has increased severely, bringing concern to policy makers since these may have some effects on price levels and/or volatility. In the first chapter, I propose an innovative structure in the model to study this. More specifically, I subdivide time periods in four quarters, and each quarter with its own specific parameters. That is, only in the first quarter there is production, and demand presents seasonal effects for each of the four quarters. My intention is to improve the accuracy of the model by introducing once more a more realistic framework. Once these adjustments are made, I will be able to decompose and quantify through simulations the different causes of prices changes.

In the second part, I incorporate an innovation into the standard theoretical sotrage model. The cornerstone of seasonally produced goods literature is the competitive storage model. Since production occurs only during one part of the year but consumption takes place all year along, inevitably storage appears as the main solution. Therefore, storage models have been widely used within the literature, with an important deal of success. However, not all aspects of the data have yet been explained. For instance, when it comes to agricultural goods, the model predicts that future contracts that deliver goods before the next harvest should not be strongly correlated with futures that deliver goods after the harvest takes place. The argument for this is that the first contracts deliver goods "from last year", whereas the latter ones deliver "this year's harvest". Since sources of supply are different, when new news regarding supply appear (for example, a harvest forecast) they should only affect the latter contracts, but not the first ones. The data shows however otherwise. Indeed, correlation between "new harvest" and "old harvest" futures contracts is positive and close to 1. This is the issue I address in the second chapter. The key element in my paper is that I assume that harvest comes in "continuously" within a relevant time interval instead of "all in one moment". This allows me to split the harvest between early and non-early parts. I show that the market equilibrium results in the early part end up being arbitraged with "old" future contracts, whereas the non-early section arbitrages with "new" ones. Therefore, the same source of supply gets sold on both type of contracts, allowing for supply induced positive correlation. I simulate the model and show this result is robust to changing parameter specifications, obtaining correlations between 0.7 to 1, as in the data. I also provide proof of the assumptions made to get this result, showing that they are highly realistic. These results are not incompatible with the main findings that have already been made, therefore it contributes to the literature by additionally explaining an unsolved puzzle.

In my third chapter, I analyze inflationary processes in major LATAM economies. More specifically, with other two coauthors we study inflation in Peru, Colombia, Brazil, Mexico and Chile for the past 18 years. We find that domestic factors such as intertia and expectations still play the biggest role. Foreign inflation however gains importance in some countries. With regarding to Phillips curve slopes, we find that these have been flattening in the last decade for most countries, that is, the cycl has a smaller effect than it used to have in previous decades when determining inflation.

Cover page of Query Language Extensions for Advanced Analytics on Big Data and their Efficient Implementation

Query Language Extensions for Advanced Analytics on Big Data and their Efficient Implementation


Advanced analytics and other Big Data applications call for query languages that can express the complex logic of advanced analytics, and are also amenable to efficient implementations providing high throughput and low latency. Existing systems such as Hadoop or Spark can now handle large amounts of data via MapReduce enabled parallelism, but they lack simple query languages that can express declaratively applications such as common graph and data mining algorithms, and the search for complex patterns in massive data sets. Fortunately, recent advances in recursive query languages and automata theory have paved way for extending widely used declarative query languages, such as SQL, to address these problems. Thus, in this dissertation, we propose two significant new extensions to the current SQL standards and demonstrate their efficient implementations. We first propose the Recursive-aggregate-SQL language, RaSQL. RaSQL queries assure a declarative formal fixpoint semantics that is guaranteed by the PreM property, while amenable to efficient recursive query evaluation techniques based on the Semi-Naive optimization for the fixpoint computation. The RaSQL is implemented on top of Apache Spark, achieving superior scalability and performance compared to the state-of-art systems such as Apache Giraph, GraphX and Myria. Then, we propose a new Weighted Search Pattern language, WSP, which extends the SQL-TS language. WSP is able to provide semantic rankings of the query results, and its implementation and optimization are guided by the theory of weighted automata.