Purpose Mixed events in obstructive sleep apnea (OSA) patients (mixed-OSA) indicate respiratory regulation instability and are essential for OSA pathogenesis and prognosis. It also shows a decreased compliance with continuous positive airway pressure (CPAP). Using predictors to identify mixed-OSA has significant clinical guidance for OSA precise diagnosis and treatment. This study aimed to establish a simple and accessible method for rapid screening of mixed-OSA, thus promoting OSA precise diagnosis. Patients and Methods A total of 907 patients with suspected OSA were screened, of which 513 OSA patients, including 344 with pure-OSA and 169 with mixed-OSA, were finally included in the study. The clinical characteristics and polysomnography (PSG) parameters of the two OSA groups were compared. Multivariate logistic regression analysis was used to investigate the factors affecting the morbidity of mixed-OSA. The receiver operating characteristic (ROC) curve was used to explore if some convenient PSG parameters can be used to predict mixed-OSA. Results About 33% of OSA patients were identified as mixed-OSA. Multivariate logistic regression analysis showed that apnea hypopnea index (AHI) and lowest oxygen saturation (LSO 2 ) were independently associated with mixed-OSA after adjusting for age, sex, body mass index (BMI), smoking, drinking, hypertension, and Epworth Sleepiness Score (ESS) (AHI: OR=1.046, 95% CI 1.032–1.060, P < 0.001; LSO 2 : OR=0.958, 95% CI 0.936–0.981, P < 0.001). ROC curve analysis showed that AHI > 47 or LSO 2 < 77% indicated mixed-OSA. The sensitivity and specificity of AHI> 47 was 0.952 and 0.652, respectively, and 0.822 and 0.675 for LSO 2 < 77%, respectively. Conclusion Our research found that AHI > 47 or LSO 2 < 77% are independently associated with mixed-OSA and can be used to quickly identify the occurrence of mixed-OSA. Therefore, this study can help detect mixed-OSA and precise individual diagnosis of OSA patients.

No abstract

In order to investigate the effects of a Chinese herbal formula Heat-stress-releasing on the antioxidant function in dairy cows, ten dairy cows were randomly divided into the control group and the experimental group, with five cows in each group. All the cows were fed with a basal diet. The animals in the experimental group were given with 220 g of herbs per day in addition to the basal diet. The trial was conducted for 14 days. Blood samples were taken from the vena cava at day 0, day 7, and day 15, respectively. The antioxidant statuses were examined. The results are as follows.(1) Heat-Stress-releasing formula can significantly increase the milk yield of dairy cows under heat stress. Compared with the control group, the milk yield of the herb-treated group increased by 14.01% (P<0.05), 14.32% (P<0.05) and 15.01% (P<0.05) in prophase, metaphase and anaphase of the test, respectively. (2) Heat-Stress-releasing formula can increase significantly the antioxidant status of the heat stressed dairy cows. Compared with the control group, the superoxide dismutase (SOD) activity increased by 45.93% (P<0.01) at day 7 and by 54.40% (P<0.01) at day 15. The Glutathioneperoxidase (GSH-PX) activity of the test group increased by 17.99% (P<0.05) at day 7 and 25.98% (P<0.01) at day 15. The total antioxidant capacity (T-AOC) of the test group increased by 43.64% (P<0.01) at day 7 and 46.35% (P<0.01) at day 15. The malondaldehyd (MDA) content of test group declined by 23.88% (P<0.01) at day 7 and 25.32% (P<0.01) at day 15.

Graph neural networks (GNN) have recently emerged as a vehicle for applying deep network architectures to graph and relational data. However, given the increasing size of industrial datasets, in many practical situations the message passing computations required for sharing information across GNN layers are no longer scalable. Although various sampling methods have been introduced to approximate full-graph training within a tractable budget, there remain unresolved complications such as high variances and limited theoretical guarantees. To address these issues, we build upon existing work and treat GNN neighbor sampling as a multi-armed bandit problem but with a newly-designed reward function that introduces some degree of bias designed to reduce variance and avoid unstable, possibly-unbounded pay outs. And unlike prior bandit-GNN use cases, the resulting policy leads to near-optimal regret while accounting for the GNN training dynamics introduced by SGD. From a practical standpoint, this translates into lower variance estimates and competitive or superior test accuracy across a several benchmarks.

Pre-trained language models have demonstrated superior performance in various natural language processing tasks. However, these models usually contain hundreds of millions of parameters, which limits their practicality because of latency requirements in real-world applications. Existing methods train small compressed models via knowledge distillation. However, performance of these small models drops significantly compared with the pretrained models due to their reduced model capacity. We propose MoEBERT, which uses a Mixture-of-Experts structure to increase model capacity and inference speed. We initialize MoEBERT by adapting the feed-forward neural networks in a pre-trained model into multiple experts. As such, representation power of the pre-trained model is largely retained. During inference, only one of the experts is activated, such that speed can be improved. We also propose a layer-wise distillation method to train MoEBERT. We validate the efficiency and effectiveness of MoEBERT on natural language understanding and question answering tasks. Results show that the proposed method outperforms existing task-specific distillation algorithms. For example, our method outperforms previous approaches by over 2% on the MNLI (mismatched) dataset. Our code is publicly available at https://github.com/ SimiaoZuo/MoEBERT.