Multivariate time series anomaly detection (MTAD) plays a vital role in a wide variety of real-world application domains. Over the past few years, MTAD has attracted rapidly increasing attention from both academia and industry. Many deep learning and graph learning models have been developed for effective anomaly detection in multivariate time series data, which enable advanced applications such as smart surveillance and risk management with unprecedented capabilities. Nevertheless, MTAD is facing critical challenges deriving from the dependencies among sensors and variables, which often change over time. To address this issue, we propose a coupled attentionbased neural network framework (CAN) for anomaly detection in multivariate time series data featuring dynamic variable relationships. We combine adaptive graph learning methods with graph attention to generate a global-local graph that can represent both global correlations and dynamic local correlations among sensors. To capture inter-sensor relationships and temporal dependencies, a convolutional neural network based on the global-local graph is integrated with a temporal self-attention module to construct a coupled attention module. In addition, we develop a multilevel encoder-decoder architecture that accommodates reconstruction and prediction tasks to better characterize multivariate time series data. Extensive experiments on real-world datasets have been conducted to evaluate the performance of the proposed CAN approach, and the results show that CAN significantly outperforms state-of-the-art baselines.

Background and objectives: The atherogenic index of plasma (AIP) is elevated in fatty liver disease, but its value in non-obese people with non-alcoholic fatty liver disease (NAFLD) is unclear. This study aimed to investigate the relationship between AIP and NAFLD as well as to determine whether AIP might be used as an indicator of NAFLD in non-obese individuals. Methods : The present study involved non-obese Chinese and Japanese participants. Univariate and multivariate logistic regression models were used to determine risk factors. The performance of risk factors was compared according to the area under the receiver operating characteristic curve. Results : In the unadjusted model, the odds ratio (OR) for every 1 standard deviation (SD) increase in AIP was 52.30. In adjusted models I and II, the OR for every 1 SD increase in AIP was 36.57 and 50.84, respectively. The area under the receiver operating characteristic curve for AIP was 0.803 and 0.802 in the development and validation groups, respectively. The best cutoff value of AIP for discrimination between NAFLD and non-NAFLD was 0.005 in the Chinese group and-0.220 in the Japanese group. Conclusions : AIP and NAFLD are positively correlated in Chinese and Japanese populations. Therefore, AIP can be used as a new screening indicator for non-obese people with NAFLD in different nations.

Precision measurements of ultra-small linear velocities of one of the mirrors in a Michelson interferometer are performed using two different weak-values techniques. We show that the technique of Almost-Balanced Weak Values (ABWV) offers practical advantages over the technique of WeakValue Amplification (WVA), resulting in larger signal-to-noise ratios and the possibility of longer integration times due to robustness to slow drifts. As an example of the performance of the ABWV protocol we report a velocity sensitivity of 60 fm/s after 40 hours of integration time. The sensitivity of the Doppler shift due to the moving mirror is of 150 nHz.

Aims/IntroductionDyslipidemia is commonly present in type 2 diabetes mellitus patients. Recently, the triglyceride : high‐density lipoprotein cholesterol (TG/HDL‐C) ratio, a novel parameter of lipid abnormality, has been seen as an independent predictor for incident diabetes. However, the correlation of the TG/HDL‐C ratio with incident diabetes in the Chinese population and how this relationship is impacted by sex have been rarely studied. In the present study, the correlation of the TG/HDL‐C ratio with incident diabetes is investigated between different sexes of the Chinese population.Materials and MethodsA total of 116,855 participants who were free of diabetes at baseline were enrolled in the study. The participants were grouped by the median value (0.82) of the TG/HDL‐C ratio. Then, participants were further analyzed according to their sex. Cumulative incidence and person‐years incidence were used to express the incidence rate. The predictive value of the TG/HDL‐C ratio for incident diabetes was probed by the Cox regression proportional hazards model.ResultsThe mean age of the participants was 44.1 ± 12.9 years, and 53% of participants (n = 62,868) were the men. A total of 2,685 incident diabetes cases occurred during the 3.1 years of the median follow‐up period. The cumulative incidence in total incident diabetes patients, men and women was 2.30% (2.21–2.38%), 3.01% (2.87–3.14%) and 1.47% (1.37–1.57%), respectively. After the adjustment of multivariate factors, the multivariate Cox regression analysis results showed that a higher TG/HDL‐C ratio was the independent predictive factor of incident diabetes in men (hazard ratio 1.30, 95% confidence interval 1.03–1.64), compared with women (hazard ratio 0.85, 95% confidence interval 0.53–1.38).ConclusionsAmong the Chinese population, the TG/HDL‐C ratio is an independent predictor for incident diabetes in male patients.