2,497 publications from this institution
Probabilistic reasoning has become a popular approach for modeling systems with uncertainty and solving for the most likely solution based on the data available. It has been successfully applied to many exciting fields and its applications are expanding. However, there has been little work on how they map to modern and future computing systems. Continued scaling of VLSI circuit technology is driving processor design towards explicitly parallel machines as energy constraints and diminishing return of instruction level parallelism limit the performance gain possible with monolithic processors. Using a reconfigurable chip multiprocessor (CMP) architecture as our evaluation platform, we characterize the performance of a set of probabilistic inference algorithms. Our results show that probabilistic inference applications have plenty of data parallelism that is easily extractable in most cases. For some applications, hardware supported multicontext processors and fine-grain synchronization are necessary to achieve efficient parallel execution. Unlike parallel scientific benchmarks, these applications have lower compute to memory ratio as well as large working sets. Thus, high memory bandwidth is required and using multi-context processors to hide some of the memory latency is desired. Besides the shift to CMP's, technology scaling is also fueling a growing concern on the reliability of future processor chips, as shrinking feature size and lower voltage make devices more susceptible to upsets from transient errors. Since probabilistic reasoning is designed to handle noisy or incomplete data, it raises the question of whether they are more robust than traditional programs, and if that warrants a different approach to soft error protection. Our fault injection experiments confirm that the robustness of approximate inference algorithms makes them more resilient against transient errors compared to traditional benchmarks. In addition, the approximate nature of the computation enables low-cost fault recovery. With simple modifications in the software, we can further improve the percentage of soft errors masked. The errors that the algorithm cannot naturally recover from often point to critical sections of the program and we find that algorithm specific software level error protection can be very effective in increasing robustness while incurring little additional overhead.
Katherine A. Grisanzio, BS; Andrea N. Goldstein-Piekarski, PhD; Michelle Yuyun Wang, BPsySc; Abdullah P. Rashed Ahmed, MS; Zoe Samara, PhD; Leanne M. Williams, PhD
Misclassification of phenotype status can seriously affect accuracy in association studies, including studies of genetic risk factors. A common problem is the classification of participants as nondiseased because of insufficient diagnostic workup or because participants have not been followed up long enough to develop disease. Some validated predictive models may have high discrimination in predicting disease. We suggest that information from such models can be used to predict the risk that a nondiseased participant will eventually develop disease and to recode the status of participants predicted to be at highest risk. We evaluate conditions under which recoding results in a maximal net improvement in the accuracy of phenotype classification. Net improvement is expected only when the positive likelihood ratio of the predictive model is larger than the inverse of the odds of disease among apparently nondiseased controls. We conducted simulations to probe the impact of reclassification on the power to detect new risk factors under several scenarios of classification accuracy of the previously developed models. We also apply this framework to a validated model of progression to advanced age-related macular degeneration that uses genetic and nongenetic variables (area under the curve = 0.915). In the training cohort (n = 2,937) and a separate validation cohort (n = 1,227), 195-272 and 78-91 nonprogressor participants, respectively, were reclassified as progressors. Correction of phenotype misclassification based on highly informative predictive models may be helpful in identifying additional genetic and other risk factors, when there are validated risk factors that provide strong discriminating ability.
No abstract is provided for this article.
Policy Points : Currently, there is massive production of unnecessary, misleading, and conflicted systematic reviews and meta‐analyses. Instead of promoting evidence‐based medicine and health care, these instruments often serve mostly as easily produced publishable units or marketing tools. Suboptimal systematic reviews and meta‐analyses can be harmful given the major prestige and influence these types of studies have acquired. The publication of systematic reviews and meta‐analyses should be realigned to remove biases and vested interests and to integrate them better with the primary production of evidence. Context Currently, most systematic reviews and meta‐analyses are done retrospectively with fragmented published information. This article aims to explore the growth of published systematic reviews and meta‐analyses and to estimate how often they are redundant, misleading, or serving conflicted interests. Methods Data included information from PubMed surveys and from empirical evaluations of meta‐analyses. Findings Publication of systematic reviews and meta‐analyses has increased rapidly. In the period January 1, 1986, to December 4, 2015, PubMed tags 266,782 items as “systematic reviews” and 58,611 as “meta‐analyses.” Annual publications between 1991 and 2014 increased 2,728% for systematic reviews and 2,635% for meta‐analyses versus only 153% for all PubMed‐indexed items. Currently, probably more systematic reviews of trials than new randomized trials are published annually. Most topics addressed by meta‐analyses of randomized trials have overlapping, redundant meta‐analyses; same‐topic meta‐analyses may exceed 20 sometimes. Some fields produce massive numbers of meta‐analyses; for example, 185 meta‐analyses of antidepressants for depression were published between 2007 and 2014. These meta‐analyses are often produced either by industry employees or by authors with industry ties and results are aligned with sponsor interests. China has rapidly become the most prolific producer of English‐language, PubMed‐indexed meta‐analyses. The most massive presence of Chinese meta‐analyses is on genetic associations (63% of global production in 2014), where almost all results are misleading since they combine fragmented information from mostly abandoned era of candidate genes. Furthermore, many contracting companies working on evidence synthesis receive industry contracts to produce meta‐analyses, many of which probably remain unpublished. Many other meta‐analyses have serious flaws. Of the remaining, most have weak or insufficient evidence to inform decision making. Few systematic reviews and meta‐analyses are both non‐misleading and useful. Conclusions The production of systematic reviews and meta‐analyses has reached epidemic proportions. Possibly, the large majority of produced systematic reviews and meta‐analyses are unnecessary, misleading, and/or conflicted.
ABSTRACT Background In the last two decades, many new interventions have been introduced with the ultimate goal of improving overall postoperative outcomes after cardiac operations in adults. We aimed to assess how often randomized controlled trials (RCTs) in adult cardiac surgery found significant mortality benefits for newer interventions versus older ones, whether observed treatment effect estimates changed over time and whether RCTs and non-randomized observational studies gave similar results. Methods We searched journals likely to publish systematic reviews on adult cardiac surgery for meta-analyses of mortality outcomes and that included at least one RCT, with or without observational studies. Relative treatment effect sizes were evaluated overall, over time, and per study design. Results 73 meta-analysis comparisons (824 study outcomes on mortality, 519 from RCTs, 305 from observational studies) were eligible. The median mortality effect size was 1.00, IQR 0.54-1.30 (1.00 among RCTs, 0.91 among observational studies, p=0.039). 4 RCTs and 6 observational studies reached p<0.005 favoring newer interventions. 2/73 meta-analyses reached p<0.005 favoring the newer interventions. Effect size for experimental interventions relative to controls did not change over time overall (p=0.64) or for RCTs (p=0.30), and there was a trend for increase in observational studies (p=0.027). In 34 meta-analyses with both RCTs (n=95) and observational studies (n=305), the median relative summary effect (summary effect in observational studies divided by summary effect in RCTs) was 0.87 (IQR, 0.55-1.29); meta-analysis of the relative summary effects yielded a summary of 0.93 (95% CI, 0.74-1.18). Conclusions The vast majority of newer interventions had no mortality differences over older ones both overall and in RCTs in particular, while benefits for newer interventions were reported more frequently in observational studies.
The foremost goal of superscalar processor design is to increase performance through the exploitation of instruction-level parallelism (ILP). Previous studies have shown that speculative execution is required for high instruction per cycle (IPC) rates in non-numerical applications. The general trend has been toward supporting speculative execution in complicated, dynamically-scheduled processors. Performance, though, is more than just a high IPC rate; it also depends upon instruction count and cycle time. Boosting is an architectural technique that supports general speculative execution in simpler, statically-scheduled processors. Boosting labels speculative instructions with their control dependence information. This labelling eliminates control dependence constraints on instruction scheduling while still providng full dependence information to the hardware. We have incorporated boosting into a trace-based, global scheduling algorithm that exploits ILP without adversely affecting the instruction count of a program. We use this algorithm and estimates of the boosting hardware involved to evaluate how much speculative execution support is really necessary to achieve good performance. We find that a statically-scheduled superscalar processor using a minimal implementation of boosting can easily reach the performance of a much more complex dynamically-scheduled superscalar processor.
No abstract is provided for this article.
Rsim is a switch-level simulator which can simulate large digital MOS integrated circuits with speedups of over three orders of magnitude over SPICE. Unfortunately, Rsim's simple switched-resistor model renders it incapable of simulating certain CMOS and most BiCMOS and ECL digital circuits. We observe that the switched-resistor model is just one particular piecewise linear model and that Rsim's simulation framework can accommodate more elaborate piecewise linear models. The resulting simulator, Mom, combines the efficiency of switch-level simulation with the ability to simulate a wider variety of circuits. We demonstrate Mom's efficiency and flexibility on a variety of circuits.
Our website uses cookies to enhance your experience. By continuing to use our site, or clicking "Continue," you are agreeing to our Cookie Policy | Continue JAMA HomeNew OnlineCurrent IssueFor Authors Publications JAMA JAMA Network Open JAMA Cardiology JAMA Dermatology JAMA Health Forum JAMA Internal Medicine JAMA Neurology JAMA Oncology JAMA Ophthalmology JAMA Otolaryngology–Head & Neck Surgery JAMA Pediatrics JAMA Psychiatry JAMA Surgery Archives of Neurology & Psychiatry (1919-1959) Podcasts Clinical Reviews Editors' Summary Medical News Author Interviews More JN Learning / CMESubscribeJobsInstitutions / LibrariansReprints & Permissions Terms of Use | Privacy Policy | Accessibility Statement 2023 American Medical Association. All Rights Reserved Search All JAMA JAMA Network Open JAMA Cardiology JAMA Dermatology JAMA Forum Archive JAMA Health Forum JAMA Internal Medicine JAMA Neurology JAMA Oncology JAMA Ophthalmology JAMA Otolaryngology–Head & Neck Surgery JAMA Pediatrics JAMA Psychiatry JAMA Surgery Archives of Neurology & Psychiatry Input Search Term Sign In Individual Sign In Sign inCreate an Account Access through your institution Sign In Purchase Options: Buy this article Rent this article Subscribe to the JAMA journal
Techniques for reducing power consumption and bandwidth limitations of inter-chip communication have been getting more attention to improve the performance of modern digital systems. This chapter begins with a brief overview of high-speed link design and describes...
