Search

Specialized accelerators are increasingly used to meet the power-performance goals of emerging applications such as machine learning, image processing, and graph analysis. Existing accelerator programming methodologies using APIs have several limitations: (1) The application code lacks portability to other platforms and compiler frameworks; (2) the lack of integration of accelerator code in the compiler limits useful optimizations such as instruction selection and operator fusion; and (3) the opacity of the accelerator function semantics limits the ability to check the final code for correctness. The root of these limitations is the lack of a formal software/hardware interface specification for accelerators. In this paper, we use the recently developed Instruction-Level Abstraction (ILA) for accelerators to serve this purpose, similar to how the Instruction Set Architecture (ISA) has been used as the software/hardware interface for processors. We propose a compiler flow termed D2A using the ILA and present a prototype that demonstrates this flow for deep learning (DL) applications. This prototype compiles programs from high-level domain-specific languages, e.g., PyTorch and MxNet, to multiple target accelerators with no target-specific extensions to the application or compiler - thus demonstrating application portability. It includes compiler optimizations through instruction selection using equality saturation-based flexible matching. Finally, we show checking the correctness of the resulting code through formal verification of individual matched operations and fully automated simulation-based validation of complete applications. The evaluation of the prototype compiler is based on six different DL applications and three different accelerators. Overall, this methodology lays the foundation for integrating accelerators in compiler flows using a formal software/hardware interface.

Additional file 1: Table S1. Mean BMI (kg/m2) in preschool children aged 4-6 years in Rhine-Neckar County and the City of Heidelberg, 2013-2018. Table S2. Trends in the prevalence of overweight in preschool children, 2013 to 2018. Table S3. Trends in the prevalence of overweight in male children, 2013 to 2018. Table S4. Trends in the prevalence of overweight in female children, 2013 to 2018. Table S5. Trends in the prevalence of obesity in preschool children, 2013 to 2018. Table S6. Trends in the prevalence of obesity in male children, 2013 to 2018. Table S7. Trends in the prevalence of obesity in female children, 2013 to 2018.

Addressing the challenge of climate change requires the large-scale development of significant renewable energy generation, but also requires these intermittent energy sources to be balanced by energy storage or demand management to maintain a reliable electric grid. In addition, a centralized generation paradigm fails to capture and utilize thermal energy for combined heat and power, abandoning a large portion of the available value from the primary energy source. A solar thermal electric system utilizing Stirling engines for energy conversion solves both of these shortcomings and has the potential to be a key technology for renewable energy generation. The ability to store thermal energy cheaply and easily allows the reliable generation of output power even during absences of solar input, and operating as distributed generation allows the output thermal stream to be captured for local heating applications. Such a system also can achieve relatively high conversion efficiencies, is fabricated using common and benign materials, and can utilize alternate sources of primary energy in an extended absence of solar input.This dissertation discusses the design, fabrication, and testing of a Stirling engine as the key component in a solar thermal electric system. In particular, the design addresses the low temperature differential that is attainable with distributed solar with low concentration ratios and is designed for low cost to be competitive in the energy space. The dissertation covers design, fabrication, and testing of a 2.5 kW Stirling Engine with a predicted thermal-to-mechanical efficiency of 20%, representing 60% of Carnot efficiency, operating between 180°C and 30°C. The design process and choices of the core components of the engine are discussed in detail, including heat exchangers, regenerator, pistons, and motor/alternator, and the process for modeling, simulation, and optimization in designing the engine. Finally, the dissertation covers the assembly and experimental testing that validates the design in terms of heat exchanger performance, losses, kinematics, and cycle work.

Additional file 1: Table S1. Descriptive statistics by season. Table S2. Descriptive statistics by quartiles of rural population. Table S3. Descriptive statistics for complete-case analysis. Table S4. Percent increase in daily CVD admissions and 95% confidence intervals (CI) per 10 µg/m3 for all PM2.5 products. Table S5. Quasi-Akaike’s Information Criterion (qAIC) for select seasonal and long-term trends. Figure S1. Percent increase in daily CVD admissions rates per interquartile range for all PM2.5 products. Figure S2. Percent increase in daily CVD admissions rates for subset excluding AQS. Figure S3. Percent increase in lag 0–1 CVD admissions rates per 10 µg/m3 for all PM2.5 products. Figure S4. Sensitivity analysis adding bank holiday as a potential confounder. Figure S5. Time series of average PM2.5 exposure estimates by PM2.5 product.

Background: Infectious diarrhea can lead to a considerable global disease burden. Thus, the accurate prediction of an infectious diarrhea epidemic is crucial for public health authorities. This study was aimed at developing an optimal random forest (RF) model, considering meteorological factors used to predict an incidence of infectious diarrhea in Jiangsu Province, China. Methods: An RF model was developed and compared with classical autoregressive integrated moving average (ARIMA)/X models. Morbidity and meteorological data from 2012 to 2016 were used to construct the models and the data from 2017 were used for testing. Results: The RF model considered atmospheric pressure, precipitation, relative humidity, and their lagged terms, as well as 1–4 week lag morbidity and time variable as the predictors. Meanwhile, a univariate model ARIMA(1,0,1)(1,0,0) 52 (AIC=−575.92, BIC=−558.14) and a multivariable model ARIMAX(1,0,1)(1,0,0) 52 with 0-1 week lag precipitation (AIC=−578.58, BIC=−578.13) were developed as benchmarks . The RF model outperformed the ARIMA/X models with a mean absolute percentage error (MAPE) of approximately 20% . The performance of the ARIMAX model was comparable to that of the ARIMA model with a MAPE reaching approximately 30%. Conclusions: The RF model fitted the dynamic nature of an infectious diarrhea epidemic well and delivered an ideal prediction accuracy . It comprehensively combined the synchronous and lagged effects of meteorological factors; it also integrated the autocorrelation and seasonality of the morbidity. The RF model can be used to predict the epidemic level and has a high potential for practical implementation.

Air pollution is one of the leading environmental problems of the 21st century, and the rise of global urbanization has increasingly exacerbated air pollution’s public health impact. Existing epidemiologic studies have tackled the relationship between air pollution and health from a variety of perspectives, but many areas of research remain lacking, including studies originating from developing countries, the assessment of exposure windows and sensitivity of modeling choices, and a better understanding of the climate change impacts on air pollution and health. In this dissertation, I address all of the issues mentioned above. Chapter 1 formally introduces the topics of this dissertation and summarizes background information on several major air pollutants. It then provides an overview of existing research on air pollution epidemiology and describes key knowledge gaps. In Chapter 2, we conduct a systematic review of the scientific literature for data on fine particulate matter (PM2.5) in China, where historical PM2.5 data are not widely available prior to 2013. Using the 574 PM2.5 measurements we identified from the literature, we detected differences in PM2.5 levels across both geographic and economic regions of China. In Chapter 3, we investigate the associations between short- and intermediate-term exposure of nitrogen dioxide (NO2) and mortality in 42 counties in China from 2013 to 2015, and find evidence of significant associations up to seven days prior to exposure. In Chapter 4, we investigate the association between PM2.5 and hospitalizations in New York State using five separate exposure datasets from 2002 to 2012. We find that despite some fluctuations in effect estimates, the majority of models yielded consistently significantly harmful associations. In Chapter 5, we utilize a global chemistry-climate model to project ozone levels in 2050 under a variety of emissions scenarios and quantify the mortality impact associated with changes in ozone concentrations between 2015 and 2050 according to each scenario. We find that under climate change alone and adherence to current legislation, ozone-related deaths would increase. However, under a best-case scenario of maximum technologically feasible reductions in emissions, over 300,000 premature deaths related to ozone can be avoided. Finally, Chapter 6 summarizes the findings of this dissertation and discusses potential directions in future research. While much work remains to be done, this dissertation work takes an important step forward in closing existing knowledge gaps in the field of air pollution epidemiology. More importantly, we hope that our work sends a strong public health message on the importance of continuing research on air pollution and health.