Bioculture Journal
Not a member yet
1309 research outputs found
Sort by
Deep Learning for Channel Coding
No full textThesis (Ph.D.)--University of Washington, 2021Wireless Communication has become a critical backbone of the information economy in the past few decades. In this rapidly improving telecommunications landscape, a crucial role is played by channel codes. Channel coding refers to the coding of information in such a way that the transmission can be robustly decoded even under noisy conditions. Progress in channel coding research has been powered by sophisticated mathematics and driven solely by human ingenuity, and therefore, progress is necessarily sporadic. For example, from convolutional code (2G) to Polar code (5G), it took several decades of research efforts to develop a new generation of channel codes. Deep learning has revolutionized a wide variety of fields and modern AI systems built using deep learning techniques are now able to surpass humans as well as human-designed algorithms.
Motivated by the success of deep learning in other fields, in this thesis, we study the role of deep learning in tackling telecommunication system design. The first part of this thesis shows that three major channel coding problems: (a) decoder design, (b) code design, and (c) feedback code design, can be automated by applying end-to-end deep supervised learning with near-optimal performance. We show surprisingly in several of these scenarios that even when the communication channels follow well studied and canonical (text-book) models, there is a significant performance improvement from deep-learning. This can be attributed not only to the ability of the learning-based methods to adapt to the channel statistics (because it is a well-known channel), but also to its ability to design sophisticated non-linear algorithms for both encoding and decoding. The second part of this thesis studies other learning paradigms such as meta-learning and federated learning (FL), which can be applied to channel coding problems to further demonstrate the versatility of neural networks.Meta learning based neural decoder shows significant efficiency on data and computation, compared to naive fine-tuning.
Finally, inspired by the strong algorithmic connection between FL personalization and meta learning, we propose a personalized FL algorithm which improves personalization significantly
Intermodal Transfers to Light Rail: Using smartcard data to estimate transfer barriers in Seattle, WA
No full textThesis (Master's)--University of Washington, 2021Transfers are a necessary inconvenience to public transit riders. They support strong hierarchical networks by connecting various local, regional, and express lines through a variety of modes. This is true in Seattle where many lines were redrawn to feed into the Link Light Rail network. Previous studies using surveys found that perceived safety, distance, and personal health were considerable predictors of transfers. This study aims to use smartcard data and generalized linear modeling to estimate which elements of transfers are commonly overcome – and which are not – among riders boarding the Link Light Rail in Seattle and its suburbs. In this process, this study also seeks to elicit any equity implications about these barriers by comparing transfer counts with the characteristics of ridership of the destination stations and origin lines. The results of this modeling suggest broad agreement with previous studies on transfers, specifically identifying distance and perceived safety as key determinants of transfers
Pier Pressure: Addressing Ecological Opportunities of Nearshore Infrastructure in Lake Washington’s Union Bay
No full textThesis (Master's)--University of Washington, 2020Along much of Seattle’s freshwater shorelines, seemingly isolated problems like erosion and shading are compounded and repeated by docks, piers, and houseboats. This results in a much bigger ecological problem: the erasure of the critical nearshore habitat that supports all life in the lake. What innovations in nearshore infrastructure design can provide multifunctional benefits for people and the environment? This design thesis considers the existing conditions of five representative zones along the University of Washington’s waterfront. Insights from restoration ecologists, engineers, local experts, and trends in aquatic infrastructure inform the design of this urban site. Pier Pressure proposes holistic solutions through a systems approach that enhances built interventions through ecological design
Molecular Computers Built from DNA Components
No full textThesis (Ed.D.)--University of Washington, 2020At the nanoscale, the ability to control spatio-temporal interactions would give us unparalleled power. DNA strand displacement systems seem to be the perfect technology to achieve control over molecular interactions, as they have the major advantage that reactions can be predicted from domain structure alone. Strand displacement technology has resulted in a myriad of dynamic devices with applications for everything from diagnostics to biomimetic manufacturing. Our goal is to build DNA strand displacement computers to achieve control over the temporal dynamics of molecules, meaning how they interact in time, and also the spatial dynamics, or how they interact in space. To build these types of molecular computers, we take inspiration from the programmability of silicon computers, which take programming languages as input. As a primitive for a molecular programming language, we look to previous work which has shown that the behavior of formal chemical reaction networks (CRNs) can be approximated using nicked double stranded DNA (ndsDNA) gates. CRNs are a natural way to think about molecular interactions and have been shown to be Turing complete, if used as a programming language. To compose complex or large DNA strand displacement CRNs, it is desirable to compose them from small, well-characterized systems, a strategy that requires quantitatively predictable reaction kinetics. However, parameter estimation of these ndsDNA gates has thus far required fitting reaction rates for every strand displacement occurrence individually and these fitted reaction rates were found to vary over more than an order of magnitude despite toehold sequences designed to have the same length and GC content. With our work on context-independent plasmid-derived gates, high quality fits can be obtained using a single reaction rate constant k for all strand displacement steps, allowing for predictable and composable kinetic behavior. Additionally, the nicked double stranded structure of the gates allows for them to be derived from plasmids as a source for highly pure DNA for use in strand displacement experiments. In our work, we use a cloning strategy that dramatically reduces recombination events, increasing the yield of functional gates that can be used for experiments, and can also be used to control stoichiometry of chemical reactions. Here, we also show that our platform using context-independent plasmid-derived gates can also be used in a spatial setting. We built a travelling wave system by putting a synthetic autocatalytic reaction in a spatial setting, demonstrating the first steps towards an autonomous and synthetic pattern formation system on a centimeter scale based on a DNA strand displacement system
Supporting data for Narenpitak, Bretherton and Khairoutdinov (2020): The Role of Multiscale Interaction in Tropical Cyclogenesis and Its Predictability in Near-Global Aquaplanet Cloud-Resolving Simulations
No full textAuthor: Pornampai Narenpitak ([email protected]).
This archive includes data used to perform analyses on tropical cyclogenesis and its predictability in: Pornampai Narenpitak, Christopher S. Bretherton, and Marat F. Khairoutdinov, 2020. The Role of Multiscale Interaction In Tropical Cyclogenesis and Its Predictability in Near-Global Aquaplanet Cloud-Resolving Simulations. Submitted to the Journal of Atmospheric Sciences. See the "Supporting_data_description.pdf" file for a listing of files in the .zip archive and a description of their use.The data included here should enable the reproduction of major analyses in Narenpitak et al. (2020, Journal of Atmospheric Sciences, submitted) about tropical cyclogenesis and its predictability. Since the near-global aquaplenet cloud-resolving simulations (NGAqua) produce large model outputs, it is impractical to archive all of the datasets. Therefore, the author archived selected tropical cyclone cases that are discussed in detail in the manuscript.The research projects were made possible by funding from Department of Energy grant DE-SC0012451, the National Science Foundation Science and Technology Center for Multi-Scale Modeling of Atmospheric Processes (CMMAP) under grant ATM-0425247, and by a scholarship from Thailand's Ministry of Higher Education, Science, Research and Innovation. Computational resources were provided by the Extreme Science and Engineering Discovery Environment (XSEDE) grant OCI-1053575
Essays on Labor and Development Economics
No full textThesis (Ph.D.)--University of Washington, 2020This dissertation consists of three separate essays on topics in labor and development economics. The first chapter examines the impact that the increased demand for foreign players in US Major League Baseball(MLB) has had on male youth’s educational attainment in the Dominican Republic. I use a triple difference strategy (DDD) and exploit the expansion of athlete visas by the US government as an exogenous source of variation. Contrary to concerns expressed by journalists and international policy researchers about the negative impacts of MLB’s overseas player development, my findings suggest an absence of meaningful negative effects, ruling out a decrease in schooling greater than 0.07 years.The second chapter studies the short-term effects of an unconditional transfer on the
labor supply in Seongnam city, Korea. In 2016, Seongnam started a “Youth Dividend”
program, which paid out gift vouchers of 1,000,000 won (USD 950) to all of its 24-year-olds.
The transfer differs from other programs in that it is explicitly unconditional and targets a
specific age. Using data from the Local Area Labor Force Survey and the synthetic control
method, I show that the unconditional transfers had no effect on the recipients’ labor supply
at neither the extensive nor intensive margin.
The third chapter focuses on the relationship between statutory work-hour reductions
and labor supply. Reducing the number of working hours and improving work-life balance
has been an important challenge for industrialized economies. In July of 2018, South Korea
lowered its maximum working hours from 68 hours a week to 52 hours. The policy reduced
the standard hours at different times according to industry and firm size. I take advantage
of this quasi-natural experiment setting to identify the impact of standard hour reductions
on working hours and employment. Using a triple difference approach, I find that female
workers in affected firms worked 3.59 hours more per week than those in the control firms,
but that there was no significant difference for male workers. My findings show no significant
relationship between work-hour reductions and job creation
Retrievals of Drizzle and Cloud Liquid Water Contents in Stratocumulus and Implications for Subgrid-scale Impacts on Model Autoconversion and Accretion Rates
No full textThesis (Ph.D.)--University of Washington, 2020Marine stratocumulus clouds (Sc) cover large areas of the Earth and have a substantial impact on the Earth’s radiative balance by reflecting copious amounts of sunlight away from the Earth and emitting longwave radiation at a temperature close to surface. Cloud and precipitation (drizzle) liquid water content (hereafter CLWC and PLWC) are two of the most important microphysical properties of Sc which directly affect radiative transfer and the hydrological cycle, as well as play a critical role in many microphysical and planetary boundary layer processes. It is thus crucial to determine CLWC and PLWC accurately. Sc in many global climate models (GCMs) are found to precipitate too frequently and too lightly which is likely due in part to the lack of information on the subgrid variability in CLWC and PLWC in the calculation of autoconversion and accretion rates. In most GCMs, the effects of subgrid variability have been either completely ignored or incorporated by multiplying the autoconversion and accretion rates (based on grid-mean values) by an enhancement factor to account for the subgrid variability. This dissertation aims to retrieve CLWC and PLWC jointly for Sc based on a millimeter wavelength radar, and to examine the nature of spatial variability in CLWC and PLWC and its impact on the autoconversion and accretion rates. In particular, we derive enhancement factors for autoconversion and accretion rates based on the radar observations, and examine how the enhancement factors change with different factors such as the length scale (size of a GCM grid) and the frequency of below-cloud precipitation. In the first part of the dissertation (Chapter 2), the CLWC and PLWC are retrieved based on a combination of retrieval techniques including a novel Doppler spectra decomposition method that separates Doppler spectra into a cloud and a precipitation component. The radar Doppler spectra data from a vertically pointing Ka-band cloud radar, along with total liquid water path from a three-channel microwave radiometer (MWR) and radiosonde measurements are used in the retrievals. These observational data in this study were collected at the U.S. Department of Energy (DOE) Atmospheric Radiation Measurement (ARM) Eastern North Atlantic (ENA) site. At the scale of a single radar volume, the uncertainty of our retrieved PLWC is about an order of magnitude. By comparing to in-situ aircraft observations, we find on average they are in a good agreement. On the scale of one day, the uncertainty in the mean CLWC is estimated to be within 30% and the systematic errors in the mean PLWC are estimated to be less than 75%. In the second part of the dissertation (Chapter 3), the variability in CLWC and PLWC and its effects on the grid-mean autoconversion and accretion rates are examined, specifically enhancement factors for autoconversion rate (E_auto) and accretion rate (E_accr). In many studies (and model implementations) enhancement factors are formulated under the assumption that variability in cloud and precipitation mixing ratio (water content divided by the air density) can be represented by a bivariate lognormal distribution with three key parameters: (i) the fractional standard deviation of the cloud-water mixing ratio, (ii) the fractional standard deviation of precipitating water mixing ratio, and (iii) the (cross) correlation coefficient (between cloud and precipitation mixing ratio). Therefore, both the enhancement factors and these three parameters are evaluated. Overall, we find that while our retrieved joint distribution is not truly a bivariate lognormal, this framework nonetheless works well given the correct values for the three key parameters. In general, we find that E_auto and E_accr increase with grid size and have a maximum when precipitation fraction is about 0.4 – 0.6 (depending somewhat on how precipitation occurrence is defined and grid size). E_auto stays relatively unchanged due to the assumption made in the retrievals that CLWC increases linearly with height in the cloud. E_accr generally decreases from cloud base to cloud top although an increase in correlation of q_c and q_p and a decrease in the magnitude of the subgrid variability of q_p have some offsetting effects. In addition, we find that E_auto and E_accr have little if any correlations with relative humidity (RH), lower tropospheric stability (LTS), and mean liquid water path (LWP) or mean cloud thickness. However, they are highly correlated with variability in of LWP, cloud thickness and cloud base, suggesting that any knowledge in subgrid variability might be useful in predicting E_auto and E_accr
High-throughput methods of studying human cytochrome P450 activity in Saccharomyces cerevisiae
No full textThesis (Ph.D.)--University of Washington, 2020Genetic variation in cytochrome P450 enzymes leads to inter-individual variability in drug metabolism, while lack of functional annotation prevents most variants from being used to inform drug choice and dosing. To measure P450 activity in a high-throughput manner, we developed a yeast-based activity assay. S. cerevisiae has a long history of use as a heterologous system of mammalian CYP expression, but we improved on this with targeted strain engineering and screening of natural isolates. In the process, we identified a sake strain with much higher human CYP expression levels than the laboratory strain. To measure P450 activity in yeast cells, we developed an assay using activity-based protein profiling with novel P450 activity-based probes to label a pooled population of variants in an activity-dependent manner. We extended this approach to a particularly important pharmacogene: CYP2C9, genetic variation in which affects the efficacy of warfarin, phenytoin, and other drugs. We coupled our yeast activity assay with fluorescence-activated cell sorting and high throughput sequencing, and generated activity scores for 6,142 single missense variants of CYP2C9. Strikingly, 65% of missense variants have significantly decreased activity suggesting altered drug metabolism in vivo. With collaborators, we also performed a second deep mutational scan of CYP2C9 in a human cell line, measuring variant abundance for 6,370 single missense variants, revealing that stability plays a large role in CYP2C9 function. Our yeast activity assay can be extended to other CYP enzymes, and will lead to advances in adverse drug response prevention by providing clinical guidance for patients carrying both currently known and yet-to-be discovered alleles of CYP2C9
Application-oriented approaches to modeling and satellite-based monitoring of watershed sediment dynamics
No full textThesis (Ph.D.)--University of Washington, 2020Global river sediment dynamics are dramatically being altered by humans in various ways, such as dam-building, deforestation, mining, and climate change. This is having major impacts on the environment and society, as sediment is vitally interconnected with water, food, energy, and ecological systems. In-situ monitoring of sediment is challenging and sediment data records are highly limited, especially in developing countries. Computational modeling and satellite remote sensing can support sediment management needs by providing sediment-related information at broad spatial and temporal scales. Considering this, researchers are frequently developing “use-inspired” tools and approaches for the modeling and monitoring of sediment dynamics, and the environment more broadly. However, use-inspired research does not always go on to become “user-ready,” or directly integrated into societal application. This is largely because engagement between researchers and stakeholders is lacking. The main objective of this dissertation is to assess how sediment dynamics research approaches of computational modeling and satellite remote sensing can be more effectively oriented towards watershed planning, management, engineering, and decision-making. First, a unique, interdisciplinary modeling approach was developed, combining state-of-the-art hydrologic and sediment transport modeling research methods. The model was tested in the Elwha River basin upstream of a former dam. The model produced accurate lifetime reservoir sedimentation volume estimates and other management-related outputs. Second, satellite remote sensing data was used to quantify changes in suspended sediment concentration (SSC) patterns due to widespread dam development in the 3S tributaries of the Mekong River Basin. Satellite data on nighttime lights and land cover helped to better explain SSC patterns. The capacities and limitations of satellite remote sensing to support sediment management were explored. Finally, a cloud-based, satellite remote sensing tool for monitoring SSC was co-developed with the Bangladesh Water Development Board. The aim of the tool was to support management of widespread riverbank erosion and riverbed accretion. The key strengths, weaknesses, and lessons learned in the stakeholder engagement process were highlighted. The two foundational strengths were a long-standing researcher-stakeholder partnership and stakeholder leadership. Overall, this dissertation demonstrates how to take sediment dynamics research tools from use-inspired to user-ready so that they effectively support growing societal needs
Where We Go and What We Carry with Us: An Autoethnographic Study of the Marginalization and Resiliency of Appalachian Indigenous Communities of West Virginia
No full textThesis (Master's)--University of Washington, 2020University of WashingtonAbstract
Where We Go and What We Carry with Us:
An Autoethnographic Study of the Marginalization and Resiliency of Appalachian Indigenous
Communities of West Virginia
Abandon GawinWaya Shuman
Chair of the Supervisory Committee:
Val Kalei Kanuha
School of Social Work
Most research claims that settler colonialism has been successful in erasing indigenous communities in West Virginia. While it is true that there are no federally recognized tribes in the state, this does not mean that there are no original indigenous people in this area. Oftentimes, the indigenous people of West Virginia are mixed with other races, including white admixtures. This study is an autoethnographical investigation into the story of my own mixed indigenous family in West Virginia. Interviews were conducted with seven of my family members from both sides of my family to provide a
narrative counterpoint about these claims. Thematic analysis of the interviews revealed three major themes: Claiming Identities: State Construction and Categorization of Race, Family as Foundation for Racial Identity; and The Role of Family: Agency in Racial/Ethnic Identity Disclosure. The results are discussed using a queer phenomenological framework to illustrate indigenous identity as more nuanced and complex than assumptions based on ideologies of the state and federal government