A5010092389- IoT and Edge/Fog Computing
- Cloud Computing and Resource Management
- Advanced Neural Network Applications
- Domain Adaptation and Few-Shot Learning
- Industrial Vision Systems and Defect Detection
- Distributed and Parallel Computing Systems
- Caching and Content Delivery
- Advancements in Battery Materials
- AI in cancer detection
- COVID-19 diagnosis using AI
- Infrastructure Maintenance and Monitoring
- Anomaly Detection Techniques and Applications
- Advanced Data Storage Technologies
- Radiomics and Machine Learning in Medical Imaging
- Traffic Prediction and Management Techniques
- Integrated Circuits and Semiconductor Failure Analysis
- Computational Drug Discovery Methods
- Age of Information Optimization
- Peer-to-Peer Network Technologies
- Advanced Battery Materials and Technologies
- Smart Grid and Power Systems
- Structural Health Monitoring Techniques
- Distributed systems and fault tolerance
- Digital Imaging for Blood Diseases
- Railway Engineering and Dynamics
Shanghai University
2022-2024
Chengdu University
2023-2024
Shanghai Electric (China)
2019-2024
Institute for Infocomm Research
2016-2024
Agency for Science, Technology and Research
2017-2024
Development Research Center
2024
State Grid Corporation of China (China)
2024
Tongji University
2019-2022
Bridge University
2022
Zhejiang Water Conservancy and Hydropower Survey and Design Institute
2009-2022
Traffic prediction is of great importance to traffic management and public safety, very challenging as it affected by many complex factors, such spatial dependency complicated road networks temporal dynamics, more. The factors make a task due the uncertainty complexity states. In literature, research works have applied deep learning methods on problems combining convolutional neural (CNNs) with recurrent (RNNs), which CNNs are utilized for RNNs dynamics. However, combinations cannot capture...
Traffic flow prediction is crucial for public safety and traffic management, remains a big challenge because of many complicated factors, e.g., multiple spatio-temporal dependencies, holidays, weather. Some work leveraged 2D convolutional neural networks (CNNs) long short-term memory (LSTMs) to explore spatial relations temporal relations, respectively, which outperformed the classical approaches. However, it hard these model jointly. To tackle this, some studies utilized LSTMs connect...
Recent graph neural network (GNN) based methods for few-shot learning (FSL) represent the samples of interest as a fully-connected and conduct reasoning on nodes flatly, which ignores hierarchical correlations among nodes. However, real-world categories may have structures, FSL, it is important to extract distinguishing features from individual samples. To explore this, we propose novel (HGNN) consists three parts, i.e., bottom-up reasoning, top-down skip connections, enable efficient...
In the recent past, security-sensitive applications, such as electronic transaction processing systems, stock quote update which require high quality of security to guarantee authentication, integrity, and confidentiality information, have adopted heterogeneous distributed system (HDS) their platforms. This is primarily due fact that single parallel-architecture-based systems may not be sufficient exploit available parallelism with running applications. Most security-aware applications end...
The increasing main memory capacity and the explosion of big data have fueled development in-memory management processing. By offering an efficient parallel execution model which can eliminate disk I/O bottleneck, existing cluster computing platforms (e.g., Flink Spark) already been proven to be outstanding for However, these are merely CPU-based systems. This paper proposes GFlink, architecture on heterogeneous CPU-GPU clusters data. Our proposed extends original from CPU clusters, greatly...
Benefiting from convenient cycling and flexible parking locations, the Dockless Public Bicycle-sharing (DL-PBS) network becomes increasingly popular in many countries. However, redundant low-utility stations waste public urban space maintenance costs of DL-PBS vendors. In this article, we propose a Bicycle Station Dynamic Planning (BSDP) system to dynamically provide optimal bicycle station layout for network. The BSDP contains four modules: drop-off location clustering, bicycle-station...
While deep learning methods hitherto have achieved considerable success in medical image segmentation, they are still hampered by two limitations: (i) reliance on large-scale well-labeled datasets, which difficult to curate due the expert-driven and time-consuming nature of pixel-level annotations clinical practices, (ii) failure generalize from one domain another, especially when target is a different modality with severe shifts. Recent unsupervised adaptation~(UDA) techniques leverage...
Predicting vehicle flows is of great importance to traffic management and public safety in smart cities, very challenging as it affected by many complex factors, such spatio-temporal dependencies with external factors (e.g., holidays, events weather). Recently, deep learning has shown remarkable performance on traditional tasks, image classification, due its powerful feature capabilities. Some works have utilized LSTMs connect the high-level layers 2D convolutional neural networks (CNNs)...
In this study, a multi-task deep neural network is proposed for skin lesion analysis. The learning model solves different tasks (e.g., segmentation and two independent binary classifications) at the same time by exploiting commonalities differences across tasks. This results in improved efficiency potential prediction accuracy task-specific models, when compared to training individual models separately. trained evaluated on dermoscopic image sets from International Skin Imaging Collaboration...
In this paper, we present an automated procedure to determine the presence of cardiomegaly on chest X-ray image based deep learning. The proposed algorithm CardioXNet uses learning methods U-NET and cardiothoracic ratio for diagnosis from X-rays. learns segmentation task ground truth data. OpenCV is used denoise maintain precision region interest once minor errors occur. Therefore, Cardiothoracic (CTR) calculated as a criterion U-net segmentations. End-to-end Dense-Net neural network...
In recent years, more and large-scale data processing computing workflow applications run on heterogeneous clouds. Such cloud with precedence-constrained tasks are usually deadline-constrained their scheduling is an essential problem faced by providers. Moreover, minimizing the execution cost based billing periods also a complex challenging for realizing this, we first model as I/O Data-aware Directed Acyclic Graph (DDAG), according to clouds global storage systems. Then, mathematically...
Intra-tumor heterogeneity (ITH) is a key challenge in cancer treatment, but previous studies have focused mainly on the genomic alterations without exploring phenotypic (transcriptomic and immune) heterogeneity. Using one of largest prospective surgical cohorts for hepatocellular carcinoma (HCC) with multi-region sampling, we sequenced whole genomes paired transcriptomes from 67 HCC patients (331 samples). We found that while ITH was rather constant across stages, had very different...
In recent years, stricter standards for lithium-ion batteries have been proposed due to the rapid development of portable electronic devices and new energy vehicles. LiNixCoyMnzO2 (NCM, x + y z = 1) has gradually become mainstream cathode materials powering its advantages high density, long cycle life, reliability. Furthermore, density NCM ternary battery is proportional nickel content. Promoting will realize their lightweight property commercial value. This research initially summarizes...
Sodium-ion batteries (SIBs) are regarded as an important substitute for lithium-ion (LIBs) due to their abundant and widespread raw material resources. The choice of the cathode has a great influence on electrochemical performance battery, Na3V2(PO4)3 (NVP) is one most promising cathodes SIBs. Its special NASICON (Na superionic conductor) three-dimensional structure conducive achieving excellent structural thermal stability during charging discharging process. Moreover, it flat...
Abstract Nickel‐rich cathode is considered to be the material that can solve short‐range problem of electric vehicles with excellent electrochemical properties and low price. However, microcracks, lithium–nickel hybridization, irreversible phase transitions during cycling limit their commercial applications. These issues should resolved by modifications. In recent years, it has been favored researchers a large number problems combining multiple modification strategies. Therefore, this paper...
As time-triggered communication protocols [e.g., controller area network (TTCAN), protocol (TTP), and FlexRay] are widely used on vehicles, the scheduling of tasks messages in-vehicle networks becomes a critical issue for offering quality-of-service (QoS) guarantees to time-critical applications vehicles. This paper studies holistic problem handling real-time in where practical aspects system design integration captured. The contributions this multifold. First, it designs novel algorithm,...
We present a deep learning framework for computer-aided lung cancer diagnosis. Our multi-stage detects nodules in 3D CAT scans, determines if each nodule is malignant, and finally assigns probability based on these results. discuss the challenges advantages of our framework. In Kaggle Data Science Bowl 2017, ranked 41st out 1972 teams.
In past years, deep convolutional neural networks (DCNN) have achieved big successes in image classification and object detection, as demonstrated on ImageNet academic field. However, There are some unique practical challenges remain for real-world recognition applications, e.g., small size of the objects, imbalanced data distributions, limited labeled samples, etc. this work, we making efforts to deal with these through a computational framework by incorporating latest developments...
Time-series forecasting is an important problem across a wide range of domains. Designing accurate and prompt algorithms non-trivial task, as temporal data that arise in real applications often involve both non-linear dynamics linear dependencies, always have some mixtures sequential periodic patterns, such daily, weekly repetitions, so on. At this point, however, most recent deep models use Recurrent Neural Networks (RNNs) to capture these which hard parallelize not fast enough for...
As an effective complement to lithium-ion batteries, sodium-ion batteries are mainly used in some low-energy-density power cells and large energy storage devices. Of the numerous anode materials for hard carbon is undoubtedly most mature currently only commercially available material, but there still a long way go before large-scale commercialization. In order reduce source cost of cathode we chose biomass as precursor batteries. this study, it demonstrated that can be classified into plant-...
Protein-protein interactions (PPIs) are crucial in various biological processes and their study has significant implications for drug development disease diagnosis. Existing deep learning methods suffer from performance degradation under complex real-world scenarios due to factors, e.g., label scarcity domain shift. In this paper, we propose a self-ensembling multi-graph neural network (SemiGNN-PPI) that can effectively predict PPIs while being both efficient generalizable. SemiGNN-PPI, not...
Abstract Mass spectrometry-coupled cellular thermal shift assay (MS-CETSA), a biophysical principle-based technique that measures the stability of proteins at proteome level inside cell, has contributed significantly to understanding drug mechanisms action and dissection protein interaction dynamics in different states. One barriers wide applications MS-CETSA is experiments must be performed on specific cell lines interest, which typically time-consuming costly terms labeling reagents mass...