Given multiple budget and revenue constraints that the transportation sector encounters, predictive analytics enables maintenance agencies to make effective decisions, prioritize maintenance tasks, and provide efficient life-cycle planning. To this end, risk-based predictive models have provided promising results in representing the susceptibility of assets to future defects. Hence, the main objective of this study is to provide an integrated framework for predicting the occurrence probability of multiple defects on different highway asset types. Several gaps in previous models were identified, including limitations in predictive frameworks given the inadequate scope of available inspection data, expert-based selection of contributing factors, and ignoring the interrelationships between neighboring assets. Therefore, this study proposes a risk-based method that combines a risk score generator and a Machine Learning (ML) algorithm to predict the hotspots of multiple defects in a given roadway. To find the best fit, the model is chosen from a pool of ML algorithms selected from different categories. To measure the efficiency of the proposed model, its performance is investigated on a selected case study. The proposed framework produced significant accurate results within the extent of available data in the case study for calculating risk scores of erosion, obstruction, and cracking on paved ditches given historical weather, traffic, maintenance, and inspection data of five selected neighboring assets (flexible pavements, unpaved ditches, slopes, small pipes and box culverts, and under drain pipes and edge drains). Additionally, the contribution of the considered factors was investigated to further study the importance of individual contributors. The framework offers decision-makers a holistic view of degradation risks of multiple assets, which could enable them to prepare an integrated asset management program. Additionally, a similar framework can be applied to other linear infrastructure systems such as sanitary sewers, water networks, and railroads.
We consider the approximation of unknown or intractable integrals using quadrature when the evaluation of the integrand is considered costly. This is a central problem in machine learning, including model averaging, (hyper-)parameter marginalization, and computing posterior predictive distributions.
Recently Batch Bayesian Quadrature (BBQ) has combined the probabilistic integration techniques of Bayesian Quadrature with the parallelization techniques of Batch Bayesian Optimization, resulting in improved performance compared to Monte Carlo techniques, especially when parallelization is increased. While the selection of batches in BBQ mitigates costs of individual point selection, every point within every batch is nevertheless chosen serially, impeding the full potential of batch selection. We resolve this shortcoming.
We developed a novel BBQ method which updates points within a batch without the costs of non-serial point selection. To implement this, we devise a dynamic domain decomposition. Combining these efficiently reduces uncertainty, lowers error estimates of the integrand, and results in more numerically robust integral estimates. Furthermore, we close an open question about the cessation criteria, which we establish and support using numerical methods.
We present our findings within the context of the history of quadrature, show how our novel methods significantly improve the literature, and provide possibilities for future research.
The essential oil (EO) industry continues to grow as consumers search for more alternative and complementary therapies. When possible, EO users are quick to turn to EOs for basic medical ailments instead of traditional medications/pharmaceuticals. With the continually high growth of EO consumers, the scientific research to support their many applications is inadequate. Due to the large gap in EO research, users do not have enough scientifically proven sources to aid in their understanding of these oils. There is a crucial need for more EO related research. A large portion of my dissertation work will provide a solid platform for users to educate themselves on EOs from a scientifically driven stand point. It will also provide new data and insights on the application and molecular mechanisms of Boswellia carterii (frankincense) EO for targeting inflammation.
The enactment of the federal G.I. Bill in 1944 and subsequent amendments over the past 76 years have provided greater access to higher education for veteran service members (Servicemen’s Readjustment Act, 1944; Steele et al., 2018; U.S. Department of Veterans Affairs, 2018a). Military-affiliated students represent the largest number of non-traditional learners entering higher education (Osborne, 2014; U.S. Department of Education, 2016; U.S. Department of Veterans Affairs, 2013; VA Campus Toolkit, 2019) with continued growth estimated in future years (VA Campus Toolkit, 2019). This current and anticipated influx of student veterans necessitates post-secondary institutions to prepare for the unique strengths, challenges, and stressors presented by student veterans in their transition from the military to college.
This phenomenological case study explored the experiences of 12 faculty and staff members in a campus-based Green Zone professional development training program intended to support the transition of student veterans into higher education. Empirical research focused on faculty and staff experiences in Green Zone training is nonexistent. Aiming to fill a void in scholarly knowledge, this study investigated how faculty and staff experienced the phenomenon of Green Zone training. The exploration was guided by four research questions: 1) What are the initial motivations of participants to engage in Green Zone training?; 2) How do faculty and staff characterize their overall experiences in the Green Zone training program?; 3) What kind of perspective changes did participants experience during the training?; and 4) What are the post-training outcomes of participants’ attendance in Green Zone training? An iterative cycle of inductive analysis yielded 12 major themes and 31 subthemes from participant narratives and triangulated by additional contextual data. Due to the interpretive nature of the study, no single theoretical framework guided the research. Instead, highlighted thematic findings were situated against theories of organizational culture and transformative learning to provide robust context to the experiences of faculty and staff in Green Zone training. Additional scholarly literature added insight to discussion of research discoveries. Findings of the study showed that organizational culture was a contributory element in participants’ overall experience in the Green Zone program, while engagement in learning that exposed them to real-life experiences of a veteran served as a pivotal point of new understanding and connection to the material. An unexpected discovery of the research was the cognitive tension that participants experienced in navigating competing ideological forces to redefine the concept of a supportive campus community for all students. Implications of this study inform application of professional development practices for higher education leaders and training practitioners in support of student veterans and other invisible and marginalized student populations.
In recent years, emotion detection in text has become more popular due to its vast potential applications in marketing, political science, psychology, human-computer interaction, artificial intelligence, etc. Access to a huge amount of textual data, especially opinionated and self-expression text, also played a special role in bringing attention to this field. In this work, we review the work that has been done in identifying emotion expressions in text and argue that although many techniques, methodologies, and models have been created to detect emotion in text, these methods, due to their handcrafted features and lexicon-based nature, are not capable of capturing the nuance of emotional language. By losing the information in the sequential nature of the text, and inability to capture the context, these methods cannot grasp the intricacy of emotional expressions, therefore, are insufficient to create a reliable and generalizable methodology for emotion detection. By understanding these limitations, we present our deep neural network methodology based on bidirectional GRU and attention mechanism and the fine-tuned transformer model (BERT) to show that we can significantly improve the performance of emotion detection models by capturing more informative text representation. Our results show a huge improvement over conventional machine learning methods on the same dataset with an average of 26.8 point increase in F-measure on the test data and a 38.6 point increase on a new dataset unseen by our model. We Show that a bidirectional-GRU with attention could perform slightly better than BERT. We also present a new methodology to create emotionally fitted embeddings and show that these embeddings perform up to 13% better in emotion similarity metrics.
Afterschool programs play a significant role in the lives of minoritized students, offering a safe space for them to develop academically, socially, and emotionally. Program administrators are responsible for the oversight of the organization and must ensure that all staff members receive the necessary professional development to impact the lives of the students and families they serve. The purpose of this qualitative study was to understand the professional development needs of afterschool and out-of-school time administrators regarding culturally relevant pedagogy. The study was framed in culturally relevant pedagogy as theorized by Gloria Ladson-Billings. A case study methodology using interview data from 5 afterschool program administrators and a document analysis addressed the three research questions. Using a thematic data analysis, three themes were derived from the data: (1) making meaning of culture; (2) seeking knowledge; and (3) enacting culturally relevant pedagogy. The findings of the study revealed that afterschool programs engage in culturally-related activities but do not institute the tenets of culturally relevant pedagogy with intent. In order to build the understanding of these paraprofessionals, culturally relevant trainings should demonstrate disparate treatment through interactive activities, offer opportunities for collaboration and include ways to link current practices to the theory of culturally relevant pedagogy. Moreover, administrators must understand the content so that they can, when necessary, deliver the training to their staff with fidelity.
Despite its long history in the United States and abroad, the unconventional drilling industry, and specifically hydraulic fracturing technology, remain controversial. While the competing demands of energy from oil and gas are contrasted with environmental safety and protection, it is likely that unconventional drilling will remain a source of social friction and a wicked problem. From the viewpoint of social resilience in hydraulic fracturing communities, social conflict represents a potential threat to the bonds that are formed within a community. This research seeks to understand the impact of planning in communities that have implemented unconventional drilling technology by using a metric of litigation as a proxy for conflict. By seeking to illuminate how conflict is affected by both municipal and industry planning efforts this research seeks to answer the question of whether planning can reduce conflict and build resilience in communities where unconventional drilling is occurring. If conflict through litigation can be reduced through planning in these communities, then resilience may be preserved, enabling these extractive communities to reduce their exposure to disruption. This research begins with a quantitative analysis of the counties in Pennsylvania to determine which counties have detailed comprehensive plans that address unconventional drilling. The comprehensive plan data was then compared to the civil lawsuit data for each county to determine which counties have both detailed comprehensive plans and low rates of fracking related civil lawsuits. Using this quantitative data, three counties were chosen as case studies for the second phase of this research. Two counties demonstrating a high level of planning and also a corresponding level of social resilience were selected (Sullivan and Clinton counties). For contrast, one county with a high level of planning, but a low level of social resilience as measured by a high incidence of civil lawsuits per well was also studied (Lawrence). A series of semi structured interviews were conducted with community members and government staff to investigate the impact of planning in those counties relative to the unconventional drilling industry. While most unconventional drilling companies declined to be interviewed for this research, one company and an industry group were also interviewed. In Sullivan County, the social resilience appears to stem from the interconnectivity of residents, government, and industry that is encouraged by the comprehensive plan and further nurtured through industry involvement in the community. In contrast, Clinton’s plan provides a guiding vision for the industry, encouraging development upon prescribed paths that promotes conscientious and environmentally and socially responsible activity. In contrast, Lawrence county’s plan addressing unconventional drilling but is stymied by a lack of reciprocal interconnectivity from industry, though the county as adapted by transitioning to related industry by leveraging their manufacturing know-how. Social resilience is notoriously difficult to measure, but this research does provide support for the theory that counties that engage in high levels of planning and also have fracking companies that are active in community engagement may have improved social resilience through the building of social bonds.
We review prior research and mathematical models examining the clearing of liabilities within financial networks, the network dynamics that affect members’ abilities to clear, and the role of financial contagion in propagating defaults across a network. Implementing the Banks as Tanks model introduced by Sonin and Sonin (2017, 2020) as a coding solution to derive a network’s clearing payment vector as defined by Eisenberg and Noe (2001), we explore clearing outcomes for a network’s members based on initial information about each’s cash and debt positions. Extending dynamics observed in the Banks as Tanks model and others, we also extend these models’ analysis of outcomes to examine the factors impacting the effectiveness of attempts to rescue defaulting members through provision of outside funding and investment. Our primary contribution is development of a framework to identify optimal interventions a regulator may impose to prevent defaults caused by a bank’s own illiquidity or by financial contagion from other defaulting banks. Secondary contributions include evaluation of the impact of network structure on intervention cost through simulations and our evaluation of methods for simplification of ergodic network or sub-network structures. Our analysis also provides a framework for further analysis of interventions within more complex networks.
Recent advances in Deep Learning have made possible distributed multi-camera IoT vision analytics targeted at a variety of surveillance applications involving automated real-time analysis of events from multiple video perspectives. However, the latency sensitive nature of these applications necessitates computing at the Edge of the network, close to the cameras. The required Edge computing infrastructure is necessarily distributed, with Cloud like capabilities such as fault tolerance, scalability, multi application tenancy, and security, while functioning at the unique operating environment of the Edge. Characteristics of the Edge include, highly heterogeneous hardware platforms with limited computational resources, variable latency wireless networks, and minimal physical security. We postulate that a distributed publish-subscribe
messaging system with storage capabilities is the right abstraction layer needed for multi-camera vision Edge analytics.
We propose Mez - a publish-subscribe messaging system for latency sensitive multi-camera machine vision at the IoT Edge. Unlike existing messaging systems, Mez allows applications to specify latency, and application accuracy bounds. Mez implements a network latency controller that dynamically adjusts the video frame quality to satisfy latency, and application accuracy requirements. Additionally, the design of Mez utilizes application domain specific features to provide low latency operations.
In this dissertation, we show how approximate computation techniques can be used to design the latency controller in Mez. We also present the design of Mez by describing its API, data model and architecture. Additionally, Mez incorporates an in-memory log based storage that takes advantage of specific features of machine vision applications to implement low latency operations. We also discuss the fault tolerance capabilities of the Mez design.
Experimental evaluation on an IoT Edge testbed with a pedestrian detection machine vision application indicates that Mez is able to tolerate latency variations of up to 10x with a worst-case reduction of 4.2% in the application inference accuracy. Further we investigated two approximate computing based algorithms - a heuristic based
pruning algorithm and a Categorical boost machine learning model based algorithm, to make the Mez’s latency controller design scalable. Both algorithms were able to achieve video frame size reduction upto 71.3% while attaining an inference accuracy of 80.9% of that of the unmodified video frames.
MICHELLE B. PASS. Staying the course: the persistence of African American biology majors at a predominantly White Institution. (Under the direction of Dr. CHANCE W. LEWIS)
Increasing the number of African Americans graduating with STEM degrees and entering the STEM workforce has been the focus of countless political reports and educational studies for decades; however, African Americans continue to experience waning graduation rates and mounting attrition rates in STEM disciplines while remaining vastly underrepresented in STEM fields. This study differs from previous studies that have focused on African Americans in STEM utilizing a deficit-based approach. This qualitative, phenomenological study examined the experiences of African American students who were successfully navigating the biology major at a predominantly White institution. This study sought to identify the factors that support the persistence of African American students in the biology major at a predominantly White institution, and to describe how these factors support their persistence in the biology major. Data were collected through in-depth interviews with six African American undergraduate biology students and analyzed using phenomenological analysis. Findings revealed that although the students were successful in the biology major, the lack of support from professors and peers within the biology major, adversely affected their academic and social experiences. Four themes emerged from the analysis of interview data. The themes are: self-determination, peer-support, independence, and adaptation. Recommendations for educational stakeholders and future research are discussed.