Wilson A.G.,CMU |
Gilboa E.,WUSTL |
Advances in Neural Information Processing Systems | Year: 2014
The ability to automatically discover patterns and perform extrapolation is an essential quality of intelligent systems. Kernel methods, such as Gaussian processes, have great potential for pattern extrapolation, since the kernel flexibly and inter-pretably controls the generalisation properties of these methods. However, automatically extrapolating large scale multidimensional patterns is in general difficult, and developing Gaussian process models for this purpose involves several challenges. A vast majority of kernels, and kernel learning methods, currently only succeed in smoothing and interpolation. This difficulty is compounded by the fact that Gaussian processes are typically only tractable for small datasets, and scaling an expressive kernel learning approach poses different challenges than scaling a standard Gaussian process model. One faces additional computational constraints, and the need to retain significant model structure for expressing the rich information available in a large dataset. In this paper, we propose a Gaussian process approach for large scale multidimensional pattern extrapolation. We recover sophisticated out of class kernels, perform texture extrapolation, inpainting, and video extrapolation, and long range forecasting of land surface temperatures, all on large multidimensional datasets, including a problem with 383,400 training points. The proposed method significantly outperforms alternative scalable and flexible Gaussian process methods, in speed and accuracy. Moreover, we show that a distinct combination of expressive kernels, a fully non-parametric representation, and scalable inference which exploits existing model structure, are critical for large scale multidimensional pattern extrapolation.
Demark-Wahnefried W.,University of Alabama at Birmingham |
Colditz G.A.,Washington University in St. Louis |
Rock C.L.,University of California at San Diego |
Sedjo R.L.,Aurora University |
And 8 more authors.
Breast Cancer Research and Treatment | Year: 2015
Obesity is a poor prognostic factor and is negatively related to quality of life (QOL) in breast cancer survivors. Exercise and Nutrition to Enhance Recovery and Good Health for You is the largest weight loss trial completed among cancer survivors. Percent losses in body weight with an intensive group-based intervention versus an attention control were 6.0 versus 1.5 % (p < 0.0001) and 3.7 versus 1.3 % (p < 0.0001) at 12 and 24 months, respectively. ENERGY also was designed to answer the research question: Does weight loss significantly improve vitality and physical function (key components of QOL)? 692 breast cancer survivors (BMI: 25–45 kg/m2) at 4 US sites were randomized to a year-long intensive intervention of 52 group sessions and telephone counseling contacts versus a non-intensive (control) of two in-person counseling sessions. Weight, self-reported QOL, and symptoms were measured semi-annually for two years. Significant decreases in physical function and increases in symptoms were observed among controls from baseline to 6 months, but not in the intervention arm, −3.45 (95 % Confidence Interval [CI] −6.10, −0.79, p = 0.0109) and 0.10 (95 %CI 0.04, 0.16, p = 0.0021), respectively. Improvements in vitality were seen in both arms but trended toward greater improvement in the intervention arm −2.72 (95 % CI −5.45, 0.01, p = 0.0508). These differences diminished over time; however, depressive symptoms increased in the intervention versus control arms and became significant at 24 months, −1.64 (95 % CI −3.13, −0.15, p = 0.0308). Increased QOL has been reported in shorter term diet and exercise trials among cancer survivors. These longer term data suggest that diet and exercise interventions improve some aspects of QOL, but these benefits may diminish over time. © 2015, Springer Science+Business Media New York.
Wright P.W.,WUSTL |
Pyakurel A.,Louisiana Tech University |
Vaida F.F.,University of California at San Diego |
Price R.W.,University of California at San Francisco |
And 11 more authors.
AIDS | Year: 2016
Objective: Little is known about the extent of cortical and subcortical volumetric alterations that may occur within the first year of HIV infection [primary HIV infection (PHI)]. Design: We used structural MRI in this prospective cross-sectional neuroimaging study to determine the extent of volumetric changes in early HIV infection. Methods: Cerebrospinal fluid, blood, neuropsychological testing, and structural T1 MRI scans were acquired from 18 HIV and 47 PHI age-matched antiretroviral-naïve male participants. Using FreeSurfer 5.1, volumetric measurements were obtained from the caudate, amygdala, corpus callosum, ventricles, putamen, thalamus, cortical white matter, and total gray matter. Regional volumes were compared groupwise and related to biomarkers in cerebrospinal fluid (viral load, neopterin, and neurofilament light chain), blood (viral load, CD4 +, and CD8 + T-cell count), and neuropsychometric tests (digit-symbol, grooved pegboard, finger-tapping, and timed gait). Results: A trend-level moderate reduction of putamen volume (P = 0.076, adjusted Cohen's d = 0.5 after controlling for age) was observed for PHI compared with HIV-uninfected individuals. Within the PHI group, putamen volume associated with CD4 + cell count (P = 0.03), CD4 + /CD8 + ratio (P = 0.045), infection duration (P = 0.009), and worsening psychomotor performance on the digit-symbol (P = 0.028), finger-tapping (P = 0.039), and timed gait (P = 0.009) tests. Conclusion: Our volumetric results suggest that the putamen is preferentially susceptible to early HIV-associated processes. Examining the natural course of early HIV infection longitudinally will allow for mapping of the trajectory of HIV-associated central nervous system changes, enabling creation of improved interventional strategies to potentially stabilize or reverse these observed structural changes. © Copyright © 2016 Wolters Kluwer Health, Inc. All rights reserved.
Mehta V.,WUSTL |
Materials Performance | Year: 2012
A study, conducted by the Washington University in St. Louis (WUSTL), was prompted by concerns that health safety programs initiated by water utilities in the United States, which use copper pipes to replace the utility's portion of lead drinking water service pipes, may actually increase the risk of lead exposure for water customers. The research shows that galvanic corrosion resulting from new copper pipes coupled with existing lead pipes that have been in service for many years can dramatically increase the amount of lead released into drinking water supplies. During galvanic corrosion, when the oxidation-reduction reaction occurs on the lead pipe, an oxidized species of elemental lead, lead +2, is formed as scale. This is the reason of the increase in the amount of lead in the Lead +2 is much more soluble than elemental lead and more readily leaches into the water. Two ways of chlorinating water to disinfect include using free chlorine, which is essentially bleach, or they use chloramines, which are basically composed of bleach combined with ammonia (NH3).