A5035966558- Advanced Bandit Algorithms Research
- Indoor and Outdoor Localization Technologies
- Advanced Causal Inference Techniques
- Statistical Methods and Inference
- Speech and Audio Processing
- Catalytic Processes in Materials Science
- Reinforcement Learning in Robotics
- Catalysis and Oxidation Reactions
- Machine Learning and Algorithms
- Statistical Methods and Bayesian Inference
- Underwater Vehicles and Communication Systems
- Lipoproteins and Cardiovascular Health
- Domain Adaptation and Few-Shot Learning
- Cardiovascular Function and Risk Factors
- Congenital Heart Disease Studies
- Cancer, Lipids, and Metabolism
- Power Line Communications and Noise
- Robotics and Sensor-Based Localization
- Functional Brain Connectivity Studies
- Advanced SAR Imaging Techniques
- Optimization and Search Problems
- Atherosclerosis and Cardiovascular Diseases
- Auction Theory and Applications
- Sparse and Compressive Sensing Techniques
- Advanced Adaptive Filtering Techniques
Beijing Children’s Hospital
2021-2025
Capital Medical University
2020-2025
Fuzhou University
2024
Fujian Institute of Research on the Structure of Matter
2024
Nanjing University of Aeronautics and Astronautics
2024
Center for Children
2023
Beijing Anzhen Hospital
2020-2023
Shanghai University of Engineering Science
2022
National University of Singapore
2021
Shanghai University
2021
We develop a deep learning framework, spatio-temporal directed acyclic graph with attention mechanisms (ST-DAG-Att), to predict cognition and disease using functional magnetic resonance imaging (fMRI). This ST-DAG-Att framework comprises of two neural networks, (1) convolutional network (ST-graph-conv) learn the spatial temporal information time series at multiple scales, where is represented by brain network, convolution over space this graph, dimension; (2) connectivity (FC-conv) features,...
Purpose Sensor arrays and pattern recognition-based electronic nose (E-nose) is a typical detection recognition instrument for indoor air quality (IAQ). The E-nose able to monitor several pollutants in the by mimicking human olfactory system. Formaldehyde concentration prediction one of major functionalities E-nose, three machine learning (ML) algorithms are most frequently used, including back propagation (BP) neural network, radial basis function (RBF) network support vector regression...
Background Although machine learning-based prediction models for in-hospital cardiac arrest (IHCA) have been widely investigated, it is unknown whether a model based on vital signs alone (Vitals-Only model) can perform similarly to that considers both and laboratory results (Vitals+Labs model). Methods All adult patients hospitalized in tertiary care hospital Japan between October 2011 2018 were included this study. Random forest with/without Vitals-Only model, respectively) trained tested...
This study aimed to investigate the effects and potential mechanisms of exercise combined with an enriched environment on learning memory in rats. Forty healthy male Wistar rats (7 weeks old) were randomly assigned into 4 groups (N = 10 each group): control (C) group, treadmill (TE) (EE) group TE + EE group. The Morris water maze (MWM) test was used evaluate ability all after eight exposure different conditions. Moreover, we employed enzyme-linked immunosorbent assay (ELISA) determine...
This study employed a deep learning longitudinal model, graph convolutional and recurrent neural network (graph-CNN-RNN), on series of brain structural MRI scans for AD prognosis. It characterized whole-brain morphology via incorporating cortical subcortical defined probabilistic risk the prediction as function age prior to clinical diagnosis. The graph-CNN-RNN model was trained half Alzheimer's Disease Neuroimaging Initiative dataset (ADNI, n = 1559) validated other ADNI Open Access Series...
Abstract Aim This study aimed to identify symptom clusters among patients with chronic heart failure (HF) and examine their independent relationships quality of life (QoL). Methods A descriptive cross‐sectional design was adopted, 201 Chinese participants were recruited. Their profiles QoL assessed using the Memorial Symptom Assessment Scale‐Heart Failure Minnesota Living Heart Questionnaire. Exploratory factor analysis used clusters. Pearson's correlation multiple regression conducted QoL....
In this paper, we describe a rotating anchor-based positioning method using time of arrival (TOA) measurements to obtain the relative positional relationship between two locators, which can help rescuer efficiently determine direction trapped person or lost find an exit in first responder applications. Since TOA ranging error special scenario cannot be modeled by distributions proposed for traditional fixed indoor geolocation, new characteristic model based on α-stable distribution is...
Homozygous familial hypercholesterolemia (HoFH) is a rare, life-threatening genetic disorder characterized by an extremely elevated serum level of low-density lipoprotein cholesterol (LDL-C) and accelerated premature atherosclerotic cardiovascular diseases (ASCVD). However, the detailed mechanism how pathogenic mutations HoFH trigger acceleration ASCVD not well understood. Therefore, we performed high-throughput RNA small sequencing on peripheral blood samples six patients three healthy...
We propose the first fully-adaptive algorithm for pure exploration in linear bandits---the task to find arm with largest expected reward, which depends on an unknown parameter linearly. While existing methods partially or entirely fix sequences of selections before observing rewards, our method adaptively changes selection strategy based past observations at each round. show sample complexity matches achievable lower bound up a constant factor extreme case. Furthermore, we evaluate...
Due to the multipath and NLOS conditions, TOA ranging error based on UWB (ultra-wide bandwidth), first path energy total received are different. In this paper, we study difference of error, in different environments, finding that is smaller than indoor scenario due metal breach environment with human body closer transmitter node, interference signal more severe.
Summary The near‐field electromagnetic ranging technology exploits the behavior of radio signals to measure distance in indoor real‐time location system. In this paper, we propose a new adaptive time‐delay estimation algorithm based on maximum correntropy criterion (MCC) and received signal strength indication (RSSI), abbreviated as RSSIMCC. proposed RSSIMCC estimates time delay between electric magnetic components low‐frequency near field instead detecting phase difference. An inaccurate...
Abstract In this research, we utilized potassium hydroxide (KOH) to re‐activate activated carbon and subsequently loaded zinc oxide obtain a more efficient catalyst for the decomposition of methyl formate. The objective KOH activation was augment catalytic activity original ZnO/AC catalyst. Various techniques were used characterize prepared samples, including BET, FT‐IR, SEM, XRD, XPS, CO 2 ‐TPD. We found that re‐activation AC can increase carrier's specific surface area alkaline sites,...
We study the problem of stochastic combinatorial pure exploration (CPE), where an agent sequentially pulls a set single arms (a.k.a. super arm) and tries to find best arm. Among variety settings CPE, we focus on full-bandit setting, cannot observe reward each arm, but only sum rewards. Although can regard CPE with feedback as special case in linear bandits, approach based bandits is not computationally feasible since number may be exponential. In this paper, first propose polynomial-time...
We consider a problem of learning binary classifier only from positive data and unlabeled (PU learning) estimating the class-prior in under case-control scenario. Most recent methods PU require an estimate probability data, it is estimated advance with another method. However, such two-step approach which first estimates class prior then trains may not be optimal since estimation error taken into account when trained. In this paper, we propose novel unified to training alternately. Our...