• Canada
• Publicationsclose
• 2017-2021close
• US close
• IL close
• AT close

Managing the thread-level parallelism (TLP) of GPGPU applications by limiting it to a certain degree is known to be effective in improving the overall performance. However, we find that such prior techniques can lead to sub-optimal system throughput and fairness when two or more applications are co-scheduled on the same GPU. It is because they attempt to maximize the performance of individual applications in isolation, ultimately allowing each application to take a disproportionate amount of shared resources. This leads to high contention in shared cache and memory. To address this problem, we propose new application-aware TLP management techniques for a multi-application execution environment such that all co-scheduled applications can make good and judicious use of all the shared resources. For measuring such use, we propose an application-level utility metric, called effective bandwidth, which accounts for two runtime metrics: attained DRAM bandwidth and cache miss rates. We find that maximizing the total effective bandwidth and doing so in a balanced fashion across all co-located applications can significantly improve the system throughput and fairness. Instead of exhaustively searching across all the different combinations of TLP configurations that achieve these goals, we find that a significant amount of overhead can be reduced by taking advantage of the trends, which we call patterns, in the way application's effective bandwidth changes with different TLP combinations. Our proposed pattern-based TLP management mechanisms improve the system throughput and fairness by 20% and 2x, respectively, over a baseline where each application executes with a TLP configuration that provides the best performance when it executes alone.

Abstract Models of aging, such as the successful aging framework outlined by Rowe & Kahn (1987; 2015) should be holistic, necessitating the inclusion of health, psychosocial factors, and social connectedness. Even at the oldest ages, life expectancy and rates of survival are increasing, yet these longer lives are accompanied with disease and disability, especially among women (Crimmins & Beltrán-Sánchez, 2010); thus, maximizing health and well-being during these post-retirement years, which can often span decades, is a high priority. However, models of age-related change, such as those relating to age-related transitions, are predominantly based on men’s experiences; less is known about how women navigate later life (Calasanti, 2010; Kim & Moen, 2002). The presentations in this symposium provide quantitative and qualitative data from women of a broad age range concerning their experiences of aging, with the shared theme of social relationships. Sherman examines the relationship between personality and social support for well-being outcomes in Native American, African American, and European American women (Mage = 57). Conceptualizing aging as the quintessential life transition, Newton outlines the diverse themes of physical, psychological, and social aging from interviews with older women (Mage = 72). Fuller and Toyama find that for older women (Mage = 80), friendships mattered more than family, and counting on neighbors could even be detrimental in terms of life satisfaction and stress. Taken together, these presentations provide a varied picture of what it means for women to ‘age well’, suggesting nuanced ways in which we might conceptualize theories of aging for women.

Partial differential equations (PDEs) are indispensable for modeling many physical phenomena and also commonly used for solving image processing tasks. In the latter area, PDE-based approaches interpret image data as discretizations of multivariate functions and the output of image processing algorithms as solutions to certain PDEs. Posing image processing problems in the infinite dimensional setting provides powerful tools for their analysis and solution. Over the last few decades, the reinterpretation of classical image processing problems through the PDE lens has been creating multiple celebrated approaches that benefit a vast area of tasks including image segmentation, denoising, registration, and reconstruction. In this paper, we establish a new PDE-interpretation of a class of deep convolutional neural networks (CNN) that are commonly used to learn from speech, image, and video data. Our interpretation includes convolution residual neural networks (ResNet), which are among the most promising approaches for tasks such as image classification having improved the state-of-the-art performance in prestigious benchmark challenges. Despite their recent successes, deep ResNets still face some critical challenges associated with their design, immense computational costs and memory requirements, and lack of understanding of their reasoning. Guided by well-established PDE theory, we derive three new ResNet architectures that fall into two new classes: parabolic and hyperbolic CNNs. We demonstrate how PDE theory can provide new insights and algorithms for deep learning and demonstrate the competitiveness of three new CNN architectures using numerical experiments. 9 pages, 4 figures, 1 table

Abstract Objective Identify the trajectories of physical and mental functioning among spouse caregivers of patients with cancer over the first five years post-diagnosis and variables associated with low or deteriorating functioning. Methods Caregivers completed a survey at 6 months and 1, 2, 3.5, and 5 years post-patient diagnosis, including the SF-12 for quality of life (QOL). SF-12 Mental Component Summary (MCS, n = 299) and Physical Component Summary (PCS, n = 300) scores were analyzed using SAS. Results Five trajectories for PCS were identified, the top three were: (a) high PCS (53.0%); (b) steady decline in PCS (17.0%); and (c) steady increase, but remaining below population norm (16.7%). Five trajectories for MCS were also identified, the top two being: (a) high MCS (45.8%) and (b) MCS comparable to population norm (27.8%). Variables associated with low or deteriorating QOL included depression, social support, coping, burden, and/or unmet needs. Conclusions This is the first study to document spouse caregivers’ QOL over the first five years post-patient diagnosis. Although many participants experienced high functioning, almost a third reported low or deteriorating mental or physical functioning. Practice implications Variables associated with low or deteriorating mental and physical functioning can be targeted in future interventions.

Inheritable and de novo variants in the cardiac voltage-gated sodium channel, Nav1.5, are responsible for both long-QT syndrome type 3 (LQT3) and Brugada syndrome type 1 (BrS1). Interestingly, a subset of Nav1.5 variants can cause both LQT3 and BrS1. Many of these variants are found in channel structures that form the channel fast inactivation machinery, altering the rate, voltage dependence, and completeness of the fast inactivation process. We used a series of mutants at position 1784 to show that the most common inheritable Nav1.5 variant, E1784K, alters fast inactivation through two separable mechanisms: (1) a charge-dependent interaction that increases the noninactivating current characteristic of E1784K; and (2) a hyperpolarized voltage dependence and accelerated rate of fast inactivation that decreases the peak sodium current. Using a homology model built on the NavPaS structure, we find that the charge-dependent interaction is between E1784 and K1493 in the DIII–DIV linker of the channel, five residues downstream of the putative inactivation gate. This interaction can be disrupted by a positive charge at position 1784 and rescued with the K1493E/E1784K double mutant that abolishes the noninactivating current. However, the double mutant does not restore either the voltage dependence or rates of fast inactivation. Conversely, a mutant at the bottom of DIVS4, K1641D, causes a hyperpolarizing shift in the voltage dependence of fast inactivation and accelerates the rate of fast inactivation without causing an increase in noninactivating current. These findings provide novel mechanistic insights into how the most common inheritable arrhythmogenic mixed syndrome variant, E1784K, simultaneously decreases transient sodium currents and increases noninactivating currents, leading to both BrS1 and LQT3. Inheritable and de novo variants of voltage-gated sodium channels that lead to sudden cardiac death can affect several different aspects of channel function, making pharmaceutical treatment difficult. Peters et al. show that the E1784K mutant in Nav1.5 alters channel function through two distinct and separable mechanisms.

Aims: The associations between potassium level and outcomes, the effect of sacubitril–valsartan on potassium level, and whether potassium level modified the effect of sacubitril–valsartan in patients with heart failure and a reduced ejection fraction were studied in PARADIGM‐HF. Several outcomes, including cardiovascular death, sudden death, pump failure death, non‐cardiovascular death and heart failure hospitalization, were examined. Methods and results: A total of 8399 patients were randomized to either enalapril or sacubitril–valsartan. Potassium level at randomization and follow‐up was examined as a continuous and categorical variable (≤3.5, 3.6–4.0, 4.1–4.9, 5.0–5.4 and ≥5.5 mmol/L) in various statistical models. Hyperkalaemia was defined as K+ ≥5.5 mmol/L and hypokalaemia as K+ ≤3.5 mmol/L. Compared with potassium 4.1–4.9 mmol/L, both hypokalaemia [hazard ratio (HR) 2.40, 95% confidence interval (CI) 1.84–3.14] and hyperkalaemia (HR 1.42, 95% CI 1.10–1.83) were associated with a higher risk for cardiovascular death. However, potassium abnormalities were similarly associated with sudden death and pump failure death, as well as non‐cardiovascular death and heart failure hospitalization. Sacubitril–valsartan had no effect on potassium overall. The benefit of sacubitril–valsartan over enalapril was consistent across the range of baseline potassium levels. Conclusions: Although both higher and lower potassium levels were independent predictors of cardiovascular death, potassium abnormalities may mainly be markers rather than mediators of risk for death.

Congenital amusia in its most common form is a disorder characterized by a musical pitch processing deficit. Although pitch is involved in conveying emotion in music, the implications for pitch deficits on musical emotion judgements is still under debate. Relatedly, both limited and spared musical emotion recognition was reported in amusia in conditions where emotion cues were not determined by musical mode or dissonance. Additionally, assumed links between musical abilities and visuo-spatial attention processes need further investigation in congenital amusics. Hence, we here test to what extent musical emotions can influence attentional performance. Fifteen congenital amusic adults and fifteen healthy controls matched for age and education were assessed in three attentional conditions: executive control (distractor inhibition), alerting, and orienting (spatial shift) while music expressing either joy, tenderness, sadness, or tension was presented. Visual target detection was in the normal range for both accuracy and response times in the amusic relative to the control participants. Moreover, in both groups, music exposure produced facilitating effects on selective attention that appeared to be driven by the arousal dimension of musical emotional content, with faster correct target detection during joyful compared to sad music. These findings corroborate the idea that pitch processing deficits related to congenital amusia do not impede other cognitive domains, particularly visual attention. Furthermore, our study uncovers an intact influence of music and its emotional content on the attentional abilities of amusic individuals. The results highlight the domain-selectivity of the pitch disorder in congenital amusia, which largely spares the development of visual attention and affective systems.

AbstractThis systematic review explored the utility of preoperative magnetic resonance imaging (MRI) as a tool for predicting intraoperative graft size for anterior cruciate ligament (ACL) reconstruction. Three databases (EMBASE, PubMed, and MEDLINE) were searched in November 2017 for English-language studies of all levels of evidence that aimed to correlate preoperative MRI measurements of common primary ACL autograft sources to intraoperative measurements of the harvested graft. Two reviewers applied predetermined inclusion/exclusion criteria to independently complete title, abstract, and full-text review of eligible studies. Data abstraction, quality assessment, and descriptive statistics are presented. A systematic screen of 930 titles resulted in 14 studies satisfying inclusion/exclusion criteria. These studies examined 762 patients of mean age 28.6 (9–67) years, with 37.3% females. Comparing the correlation of preoperative MRI measurements to intraoperative harvested measures, the strength was very highly positive for quadriceps tendon (QT) (one study, 29 patients, intraclass correlation coefficient [ICC] = 0.96), highly positive for patellar tendon (two studies, 28 patients, ICC: 0.77–0.87), negligible-highly positive for semitendinosus-only tendon (eight studies, 439 patients, r: 0.16–0.81), and negligible-moderately positive for gracilis-only tendon (four studies, 143 patients, r: 0.29–0.59). When combined semitendinosus–gracilis tendon grafts were considered, the correlation ranged from low-very highly positive (10 studies, 517 patients, r: 0.42–0.93). Preoperative MRI assessment of both QT and bone–patellar tendon–bone autografts most highly correlates with intraoperative measurements of autograft diameter. Considerable variability exists when viewing hamstring tendons either individually or together, where most studies indicate at least a moderate correlation. This highlights the advantage of MRI during the preoperative planning process in equipping the surgeon with a better ability to ensure the diameter of the intended autograft will suffice. This is a Level IV study, systematic review of Levels II to IV studies.

High cellularity and poorly organized tumour vasculature in high-grade gliomas leads to insufficient blood supply, hypoxic areas, and ultimately to the formation of necrosis. Thus, hypoxia is a hallmark of malignant glioma microenvironment and it is associated with aggressive tumor behavior such as growth, progression, and resistance to chemo-radiation. Current pathologic markers are insufficient to identify patients that may benefit from specific treatments. We therefore, hypothesized that underlying epigenetic alterations confer therapeutic resistance under hypoxic conditions. Twenty five GBM patients were consented and treated with pimonidazole (PIMO) 16–18 hours prior to surgery. Tumor sections were subjected to immunohistochemical analysis using antibodies against PIMO and other hypoxia markers such HIF1a and CAIX. Samples were subjected to laser capture microdissection followed by DNA isolation and DNA methylation profiling using the Illumina Human Methylation EPIC Array. Data was analyzed using minfi and conumee packages in Bioconductor, together with appropriate biostatistics tools. PIMO score was determined to range from 10–60% and positively correlated with other hypoxia markers such as CA IX and HIF1a (p4,000) were hypomethylated. Gene set enrichment analysis (GSEA) indicated that the majority of these CpGs are associated with genes involved in signalling cascades and oncogenic processes, including WNT and NOTCH. These were compared to DNA methylation profiles of glioma stem cells exposed to transient hypoxia and extensive overlap was found in proportion of hypomethylated CpG sites and cellular processes that were altered. These findings were correlated with complementary RNA expression data from RNA sequencing to establish the biological relevance of changes in DNA methylation profiles under hypoxia in GBM.

10 Background: For men with mCRPC, systemic therapies such as docetaxel and cabazitaxel improve survival, but more effective treatments are needed. KEYNOTE-365 (NCT02861573) is a phase 1b/2 study to examine the safety and efficacy of pembro in combination with 4 different study medications (cohorts A, B, C, D) in mCRPC. Previous data from cohort B with a median of 20 months of follow-up showed that pembro + docetaxel and prednisone was well tolerated and had antitumor activity in patients (pts) with mCRPC previously treated with abi or enza. New efficacy and safety data after an additional year of follow-up are presented. Methods: Cohort B enrolled pts who did not respond to or were intolerant to ≥4 weeks of abi or enza in the prechemotherapy mCRPC state and whose disease progressed within 6 months of screening (determined by PSA progression or radiologic bone/soft tissue progression). Pts received pembro 200 mg IV every 3 weeks (Q3W), docetaxel 75 mg/m2 IV Q3W, and oral prednisone 5 mg twice daily. Primary end points were safety, PSA response rate (PSA decrease >50% from baseline), and ORR per RECIST v1.1 by blinded independent central review. Efficacy and safety were assessed in all pts as treated. Results: Of the 104 treated pts, median age was 68.0 years (range, 50-86), 23.1% had PD-L1–positive tumors (combined positive score ≥1), 25.0% had visceral disease, and 50.0% had measurable disease. Median time from enrollment to data cutoff was 32.4 months (range 13.9-40.3); 101 pts discontinued, primarily because of disease progression (77.9%). Efficacy outcomes are reported in the table below. Treatment-related adverse events (TRAEs) occurred in 100 pts (96.2%); the most frequent (≥30%) were diarrhea (41.3%), fatigue (41.3%), and alopecia (40.4%). Grade 3-5 TRAEs occurred in 46 pts (44.2%). Five pts (4.8%) died of AEs; 2 were treatment-related pneumonitis. Conclusions: After another year of follow-up, pembro + docetaxel and prednisone showed improved ORR and PSA response rates compared to the prior dataset in pts with mCRPC previously treated with abi or enza. Safety was consistent with known profiles of each agent and will be further evaluated in a phase 3 study (KEYNOTE-921). Clinical trial information: NCT02861573. [Table: see text]