- Topic Modeling
- Authorship Attribution and Profiling
- Natural Language Processing Techniques
- Explainable Artificial Intelligence (XAI)
- Academic integrity and plagiarism
- Advanced Computational Techniques and Applications
- Advanced Text Analysis Techniques
- Machine Learning in Materials Science
- Ocean Waves and Remote Sensing
- Oceanographic and Atmospheric Processes
- Geophysics and Gravity Measurements
- Bioinformatics and Genomic Networks
- Spam and Phishing Detection
- Remote Sensing and Land Use
- Machine Learning in Healthcare
- Image Processing and 3D Reconstruction
- Generative Adversarial Networks and Image Synthesis
- Hate Speech and Cyberbullying Detection
- Advanced Algorithms and Applications
- Computer Graphics and Visualization Techniques
- Online and Blended Learning
- Marine and coastal ecosystems
- Digital Humanities and Scholarship
- Peer-to-Peer Network Technologies
- Infrared Thermography in Medicine
Anhui Agricultural University
2018-2024
Shenyang Jianzhu University
2022
Ocean University of China
2003-2021
Qingdao National Laboratory for Marine Science and Technology
2017-2021
Foshan University
2020-2021
Lawrence Livermore National Laboratory
2019-2021
Jiangxi University of Traditional Chinese Medicine
2020
CITIC Group (China)
2015
Heilongjiang Institute of Technology
2014
University of Finance and Economics
2010
WebVR, a multi-user online virtual reality engine, is introduced. The main contributions are mapping the geographical space and to P2P overlay network space, dividing three spaces by quad-tree method. geocoding identified with Hash value, which used index user list, terrain data, model object data. Sharing of data through improved Kademlia designed implemented. In this model, XOR algorithm calculate distance space. greatly improves hit rate 3D geographic search under network. Some...
In this work, we propose an introspection technique for deep neural networks that relies on a generative model to instigate salient editing of the input image interpretation. Such modification provides fundamental interventional operation allows us obtain answers counterfactual inquiries, i.e., what meaningful change can be made in order alter prediction. We demonstrate how reveal interesting properties given classifiers by utilizing proposed approach both MNIST and CelebA dataset.
In this work, we propose an introspection technique for deep neural networks that relies on a generative model to instigate salient editing of the input image interpretation. Such modification provides fundamental interventional operation allows us obtain answers counterfactual inquiries, i.e., what meaningful change can be made in order alter prediction. We demonstrate how reveal interesting properties given classifiers by utilizing proposed approach both MNIST and CelebA dataset.
To address problems of serious loss details and low detection definition in the traditional human motion posture algorithm, a algorithm using deep reinforcement learning is proposed. Firstly, perception ability used to match feature points obtain features. Secondly, normalize image, take color histogram distribution as antigen, search region close its candidate antibody. By calculating affinity between antigen antibody, extraction realized. Finally, training characteristics network network,...
The Southern Ocean front (SOF) is an important factor that affects the heat exchange and material transport of Ocean. In past two decades, with advancements in satellite remote-sensing technology, study spatio-temporal variability has become a new hot topic. Nevertheless, southwestern Atlantic, as part Ocean, poorly studied this regard. Based on 16-year (2004–2019) high-resolution observations sea surface temperature (SST) 13-year (2007–2019) chlorophyll (CHL), detected analyzed position...
Airborne Light Detection and Ranging (LiDAR) is widely used in digital elevation model (DEM) generation. However, the very large volume of LiDAR datasets brings a great challenge for traditional serial algorithm. Using parallel computing to accelerate efficiency DEM generation from points has been hot topic geo-computing. Generally, most existing algorithms running on high-performance clusters (HPC) were process-paralleling mode, with static scheduling strategy. The strategy would not...
Formal learning and informal have always been two ways of learning, but little research has done on learning. It wasn't until the establishment MOOC in 2012 that this kind off-campus gained traction. This paper first discusses definition its advantages, then introduces mobile device-based methods, makes an outlook their prospects problems.
Paraphrase identification is central to many natural language applications. Based on the insight that a successful paraphrase model needs adequately capture semantics of objects as well their interactions, we present deep interacting with syntax (DPIM-ISS) for identification. DPIM-ISS introduces linguistic features manifested in syntactic produce more explicit structures and encodes semantic representation sentence different by means syntax. Then, learns pattern from this exploiting...
This paper introduces an ancient ruin system which can be used to organize and present ruins data, providing 3D spatial analysis, visualization, interactive roaming path animation. In order perform real-time rendering of massive we a novel method compute layout large triangle meshes. The is successfully in "six-horse carriage for the emperor" horse chariot pits simulation.
Identifying various types of aluminum wheels was traditionally performed manually, which could result in low efficiency, limited reliability, poor accuracy, and high labor cost.This paper presents the design implementation an on-line automatic recognition system for based on laser trigonometry principle.The mainly consists a station, two displacement sensors, ball-screw, industrial computer with data processing software, programmable logic controller (PLC).Robust algorithms, as well...
Graph neural network (GNN) explanations have largely been facilitated through post-hoc introspection. While this has deemed successful, many explanation methods shown to fail in capturing a model's learned representation. Due problem, it is worthwhile consider how one might train model so that more amenable analysis. Given the success of adversarial training computer vision domain models with reliable representations, we propose similar paradigm for GNNs and analyze respective impact on...
With the increasing attention on protection of intellectual property rights, a large number patents need to be processed. However, since patent is kind complicated technical text, it difficult understand patents. How quickly by computer problem. To solve above problem, our method tag issue sentences, these sentences describe problems solved in Tagging very important research topic understanding, because revolves around are key patent. There two challenges task: (1) extract get corpus? (2)...