A5101583267- Human Mobility and Location-Based Analysis
- Urban Transport and Accessibility
- Multimodal Machine Learning Applications
- Water Quality Monitoring and Analysis
- Advanced Image and Video Retrieval Techniques
- Traffic Prediction and Management Techniques
- Natural Language Processing Techniques
- Remote Sensing and Land Use
- Impact of Light on Environment and Health
- Transportation Planning and Optimization
- Topic Modeling
- Human Pose and Action Recognition
- Data Management and Algorithms
- Land Use and Ecosystem Services
- Speech and dialogue systems
- Geographic Information Systems Studies
- Generative Adversarial Networks and Image Synthesis
- Pharmacological Effects of Natural Compounds
- Video Analysis and Summarization
- Hydrology and Watershed Management Studies
- Automated Road and Building Extraction
- Odor and Emission Control Technologies
- Anaerobic Digestion and Biogas Production
- Advanced Vision and Imaging
- Remote-Sensing Image Classification
Shenzhen University
2020-2025
Shenzhen Bay Laboratory
2023-2025
Chinese University of Hong Kong, Shenzhen
2024-2025
Northeastern University
2025
China-Japan Friendship Hospital
2025
Chinese Research Academy of Environmental Sciences
2016-2024
Guangdong-Hongkong-Macau Joint Laboratory of Collaborative Innovation for Environmental Quality
2022-2024
University of Hong Kong
2024
Ministry of Natural Resources
2023
Tongji University
2023
Understanding dynamic human mobility changes and spatial interaction patterns at different geographic scales is crucial for assessing the impacts of non-pharmaceutical interventions (such as stay-at-home orders) during COVID-19 pandemic. In this data descriptor, we introduce a regularly-updated multiscale flow dataset across United States, with starting from March 1st, 2020. By analysing millions anonymous mobile phone users' visits to various places provided by SafeGraph, daily weekly...
Integrating raw Global Position System (GPS) trajectories with a road network is often referred to as map-matching problem. However, low-frequency (e.g., one GPS point for every 1-2 min) have raised many challenges existing methods. In this paper, we propose novel and global spatial-temporal method called spatial temporal conditional random field (ST-CRF), which based on insights relating to: 1) the positioning accuracy of points topological information underlying network; 2) accessibility...
Fine-grained prediction of urban population is great practical significance in many domains that require temporally and spatially detailed information. However, fine-grained modeling has been challenging because the highly dynamic its mobility pattern complex space time. In this study, we propose a method to predict at large spatiotemporal scale city. This models temporal dependency by estimating future inflow with current spatial correlation using an artificial neural network. With dataset...
Trajectory prediction plays an important role in supporting many advanced applications such as location-based services and intelligent traffic managements. Most existing trajectory methods employed fixed spatial division focused on human closeness movement patterns. However, these could lead to a sharp boundary limitation ignore the periodic characteristics of mobility. This paper proposes novel method based long short-term memory network (LSTM) called predictor with fuzzy-long...
(2021). Prediction of human activity intensity using the interactions in physical and social spaces through graph convolutional networks. International Journal Geographical Information Science: Vol. 35, Spatial Social Networks GIScience, pp. 2489-2516.
Searching for a parking spot in metropolitan areas is great challenge, especially highly populated such as downtown districts and job centres. On-street often cost-effective choice compared to facilities garages lots. However, limited space complex regulation rules make the search process of on-street legal very difficult. To this end, we propose data-driven framework understanding predicting spatiotemporal legality using NYC tickets open data, points interest (POI) data human mobility data....
The spatiotemporal variability in air pollutant concentrations raises challenges linking pollution exposure to individual health outcomes. Thus, understanding the patterns of human mobility plays an important role epidemiology and studies. With advantages massive users, wide spatial coverage passive acquisition capability, mobile phone data have become emerging source for compiling estimates. However, compared with monitoring data, temporal granularity is not high enough, which limits...
In the UK, all domestic COVID-19 restrictions have been removed since they were introduced in March 2020. After illustrating spatial-temporal variations infection rates across London, this study then particularly aimed to examine relationships of with building attributes, including density, type, age, and use, previous studies shown that built environment plays an important role public health. Multisource data from national health services London Geomni map processed GIS techniques...
Hydro-Morphological Processes (HMP, any natural phenomenon contained within the spectrum defined between debris flows and flash floods) are most likely to occur in small catchments, especially buffer zones along or near rivers. Rivers transfer matter energy hydrographic units, thus potentially affecting occurrence of HMPs nearby catchments. To date, previous HMP susceptibility studies based on data-driven modeling lacked taking into account these interactions In this work, we fully...
Differing from the conventional communication system paradigm that models information source as a sequence of (i.i.d. or stationary) random variables, semantic approach aims at extracting and sending high-level features content deeply contained in source, thereby breaking performance limits statistical theory. As pioneering work this area, deep learning-enabled (DeepSC) constitutes novel algorithmic framework based on transformer--which is learning tool widely used to process text...
Existing preference optimization objectives for language model alignment require additional hyperparameters that must be extensively tuned to achieve optimal performance, increasing both the complexity and time required fine-tuning large models. In this paper, we propose a simple yet effective hyperparameter-free algorithm alignment.We observe promising performance can achieved simply by optimizing inverse perplexity, which is calculated as of exponentiated average log-likelihood chosen...
While MLLMs perform well on perceptual tasks, they lack precise multimodal alignment, limiting performance. To address this challenge, we propose Vision Dynamic Embedding-Guided Pretraining (VDEP), a hybrid autoregressive training paradigm for MLLMs. Utilizing dynamic embeddings from the MLP following visual encoder, approach supervises image hidden states and integrates tokens into training. Existing primarily focused recovering information textual inputs, often neglecting effective...
Real-time speech interaction, serving as a fundamental interface for human-machine collaboration, holds immense potential. However, current open-source models face limitations such high costs in voice data collection, weakness dynamic control, and limited intelligence. To address these challenges, this paper introduces Step-Audio, the first production-ready solution. Key contributions include: 1) 130B-parameter unified speech-text multi-modal model that achieves understanding generation,...
In our quest to decode the visual processing of human brain, we aim reconstruct dynamic experiences from brain activities, a task both challenging and intriguing. Although recent advances have made significant strides in reconstructing static images non-invasive recordings, translation continuous activities into video formats has not been extensively explored. Our study introduces NeuralFlix, simple but effective dual-phase framework designed address inherent challenges decoding fMRI data,...