A5000601581- Hydrocarbon exploration and reservoir analysis
- Recommender Systems and Techniques
- Seismic Imaging and Inversion Techniques
- Topic Modeling
- Autonomous Vehicle Technology and Safety
- Privacy-Preserving Technologies in Data
- Geological and Geophysical Studies
- Traffic control and management
- Geological Modeling and Analysis
- Natural Language Processing Techniques
- Machine Learning and ELM
- Image and Video Quality Assessment
- Mobile Health and mHealth Applications
- Advanced Text Analysis Techniques
- Machine Fault Diagnosis Techniques
- Karst Systems and Hydrogeology
- Vehicle Dynamics and Control Systems
- 3D Surveying and Cultural Heritage
- Artificial Intelligence in Healthcare
- Robotics and Automated Systems
- Expert finding and Q&A systems
- Human Mobility and Location-Based Analysis
- Energy Load and Power Forecasting
- Geoscience and Mining Technology
- Imbalanced Data Classification Techniques
Shanghai University of Electric Power
2020-2022
Nanjing University of Aeronautics and Astronautics
2021-2022
Shenzhen University
2018-2021
Research Institute of Petroleum Exploration and Development
2006-2021
Beihang University
2021
Embedded Systems (United States)
2020
Zhejiang University of Technology
2014
Chongqing University
2002-2012
Ohio University
2010
Shanghai Jian Qiao University
2006
Federated learning is a new machine paradigm which allows data parties to build models collaboratively while keeping their secure and private. While research efforts on federated have been growing tremendously in the past two years, most existing works still depend pre-existing public datasets artificial partitions simulate federations due lack of high-quality labeled generated from real-world edge applications. Consequently, advances benchmark model evaluations for lagging behind. In this...
In order to solve the accuracy problem of future motion prediction surrounding vehicles with different types drivers, this paper proposes a comprehensive lateral method that combines driver intention and vehicle behavior recognition. For optimization models are established: personal optimal model system model. Then, driver’s probability is obtained through game theory. Different from traditional theory, we apply for on premise getting type instead using Nash equilibrium all players....
Wellness data generated by patients using smart phones and portable devices can be a key part of Personal Health Record (PHR) offers healthcare service providers (healthcare providers) patient health information on daily basis. Prior research has identified the potential for improved communication between provider patient. However practice sharing wellness not been widely adopted sector; one reasons being lack interoperability preventing successful integration such device into PHR Electronic...
In order to make safe and reasonable decisions in some high-risk environments such as the mandatory lane change, we propose an IMM-based partially observable Markov decision process (POMDP) algorithm using collision-risk function which combines time-to-collision (TTC), intervehicular time (IT), collision for change. The newly proposed contains two parts: vehicle impact factor function, is used assess risk determines whether autonomous collides with surrounding vehicles. IMM-base POMDP...
Abstract Equipment usually breaks down suddenly and irregularly, so most of the data sets obtained for fault diagnosis have unbalanced characteristics, amount varies greatly from different types. In this paper, three problems in application synthetic minority oversampling technique (SMOTE) are studied, improved SMOTE algorithm combined with support vector machine (SVM) is proposed. The validity model verified by CWRU bearing compared SVM SMOTE+SVM methods, result satisfactory.
With the increasing difficulty of exploration and development in conventional oil gas resources, unconventional resources will play an important role. Especially low permeability extra reservoirs, proportion proven reserves is more large. The content brittleness minerals tight reservoirs determines effect fracturing modification, directly affects production. At present, prestack seismic data, such as Young's modulus Poisson's ratio often used on prediction sweet-spots. And predicted results...
The increasing availability of heterogeneous ambient sensing systems challenges the according information processing to analyse and compare a variety different in single scenario. For instance, localization objects can be performed by image as well radio based localization. If such are utilized localize same objects, synergy outputs is important enable comparable meaningful analysis. This demo showcases practical deployment an example system.
The 3D visualization of the traditional power plant can integrate information in scene and manage each area accurately quickly. surface missing parts holes reconstruction process lead to loss information. To solve this problem, an efficient repairing algorithm is proposed paper. Through feature discrimination original data simplify processing, then repair hole. In case retaining characteristics data, subsequent efficiency improved by reducing number point clouds. After fast triangulation...
Due to the difference of quality seismic data and complexity sequence, stratal slices will show local inconsistent phenomenon in deposit time, which affect authenticity deposition analysis. With nonlinear algorithm, shape slice horizon is recalculated depicted longitudinal direction. The new as close possible time interface. this we can extract with isochronal significance, provides a more accurate basis for sedimentation Isochronal stratigraphy technique key research method sedimentology,...
In this paper, we study an important recommendation problem with heterogeneous feedback of users' grade scores such as 5-star and like/dislike binary ratings assigned to items. As a response, address the from transfer learning perspective, i.e., taking target data auxiliary data, in order share knowledge between two different types more sufficiently. Technically, besides observed explicit ratings, propose exploit implicit preference context beneath feedback, which is incorporated into...
In order to cope with the problem of low efficiency and slow dynamic response DAB converter under traditional PI control, this paper proposes a current stress optimization(CSO) method used in design based on model predictive control. Firstly, relationship between inductor shift ratio each operating mode dual-phase-shift(DPS) modulation is analyzed, then prediction established output voltage converter, optimal input at moment derived using Lagrange multiplier basis cost function. The final...
Dolomite is one of the rocks favorable for oil and gas accumulation in marine strata, but its distribution pattern difficult to predict by geophysics. Many scholars have proposed research methods including attribute lithological inversion study dolomite, some wave clustering methods. Among them, waveform classification, existing method usually uses equal-thickness time windows select target layer analysis. For areas where thickness reservoir varies greatly vertical horizontal directions,...
The steam pipe network is a complex nonlinear structure system with large time delay, and the accurate prediction of its pipeline pressure prerequisite for energy dispatch guarantee. Considering complicated changes in many influencing factors, this paper proposes model based on CNN BiLSTM attention mechanism. This method can strengthen connection between various users, reduce loss historical information, achieve higher-precision prediction. Compared other models, MAPE reduced to 0.55%, RMSE...