Proceedings of the National Academy of Sciences of the United States of America
-
Proc. Natl. Acad. Sci. U.S.A. · Aug 2019
Effects of policy-driven hypothetical air pollutant interventions on childhood asthma incidence in southern California.
Childhood asthma is a major public health concern and has significant adverse impacts on the lives of the children and their families, and on society. There is an emerging link between air pollution, which is ubiquitous in our environment, particularly in urban centers, and incident childhood asthma. Here, using data from 3 successive cohorts recruited from the same 9 communities in southern California over a span of 20 y (1993 to 2014), we estimated asthma incidence using G-computation under hypothetical air pollution exposure scenarios targeting nitrogen dioxide (NO2) and particulate matter <2.5 μm (PM2.5) in separate interventions. ⋯ For example, compliance with a hypothetical standard of 20 ppb NO2 was estimated to result in 20% lower childhood asthma incidence (95% CI, -27% to -11%) compared with the exposure that actually occurred. The findings for hypothetical PM2.5 interventions, although statistically significant, were smaller in magnitude compared with results for the hypothetical NO2 interventions. Our results suggest a large potential public health benefit of air pollutant reduction in reduced incidence of childhood asthma.
-
Proc. Natl. Acad. Sci. U.S.A. · Aug 2019
Reconciling modern machine-learning practice and the classical bias-variance trade-off.
Breakthroughs in machine learning are rapidly changing science and society, yet our fundamental understanding of this technology has lagged far behind. Indeed, one of the central tenets of the field, the bias-variance trade-off, appears to be at odds with the observed behavior of methods used in modern machine-learning practice. The bias-variance trade-off implies that a model should balance underfitting and overfitting: Rich enough to express underlying structure in data and simple enough to avoid fitting spurious patterns. ⋯ This "double-descent" curve subsumes the textbook U-shaped bias-variance trade-off curve by showing how increasing model capacity beyond the point of interpolation results in improved performance. We provide evidence for the existence and ubiquity of double descent for a wide spectrum of models and datasets, and we posit a mechanism for its emergence. This connection between the performance and the structure of machine-learning models delineates the limits of classical analyses and has implications for both the theory and the practice of machine learning.