A5064754627- Data Management and Algorithms
- Traffic Prediction and Management Techniques
- Data Visualization and Analytics
- Transportation and Mobility Innovations
- Geographic Information Systems Studies
- Human Mobility and Location-Based Analysis
- Urban and Freight Transport Logistics
- Remote Sensing in Agriculture
- Music and Audio Processing
- Migration, Refugees, and Integration
- Automated Road and Building Extraction
- Urban Transport and Accessibility
- Time Series Analysis and Forecasting
- Video Surveillance and Tracking Methods
- Migration, Health and Trauma
- Music Technology and Sound Studies
- Advanced Manufacturing and Logistics Optimization
- Land Use and Ecosystem Services
- Migration and Labor Dynamics
- Music History and Culture
IT University of Copenhagen
2020-2024
University of Edinburgh
2018-2020
We present PyTorch Geometric Temporal, a deep learning framework combining state-of-the-art machine algorithms for neural spatiotemporal signal processing. The main goal of the library is to make temporal geometric available researchers and practitioners in unified easy-to-use framework. Temporal was created with foundations on existing libraries eco-system, streamlined network layer definitions, snapshot generators batching, integrated benchmark datasets. These features are illustrated...
Searching for similar GPS trajectories is a fundamental problem that faces challenges of large data volume and intrinsic complexity trajectory comparison. In this paper, we present suite sketches drastically reduce the computation costs associated with near neighbor search, distance estimation, clustering classification, subtrajectory detection. Apart from summarizing dataset, our have two uses. First, obtain simple provable locality sensitive hash families both Hausdorff Fréchet measures,...
Light goods vehicles (LGV) used extensively in the last mile of delivery are one leading polluters cities. Cargo-bike logistics and Electric Vehicles (LEVs) have been put forward as a high impact candidate for replacing LGVs. Studies estimated over half urban van deliveries being replaceable by cargo-bikes, due to their faster speeds, shorter parking times more efficient routes across However, sector suffers from lack publicly available data, particularly pertaining cargo-bike deliveries,...
<title>Abstract</title> During asylum seeking procedures, refugees are required to provide significant amounts of private information. When application records summarised and made publicly available or shared almost as-is for research purposes, only names typically anonymized. It is well understood, however, that mobility patterns can easily single out individuals; a few spatiotemporal points needed, even in large datasets. In this paper we show the re-identification effect from locations...
We consider the problem of finding large scale mobility patterns. A common challenge in tracking systems is that quantity data spread out spatially and temporally across many sensors. thus devise a spatial sampling information exchange protocol provides probabilistic guarantees on detecting prominent For this purpose, we define general notion significant popular paths can capture different types motion. design summary sketch for at each node, which be updated efficiently, then aggregated...
Musical genres are inherently ambiguous and difficult to define. Even more so is the task of establishing how relate one another. Yet, genre perhaps most common effective way describing musical experience. The number possible classifications (e.g. Spotify has over 4000 tags, LastFM 500,000 tags) made idea manually creating music taxonomies obsolete. We propose use hyperbolic embeddings learn a general taxonomy by inferring continuous hierarchies directly from co-occurrence large dataset....
Light goods vehicles (LGV) used extensively in the last mile of delivery are one leading polluters cities. Cargo-bike logistics has been put forward as a high impact candidate for replacing LGVs, with experts estimating over half urban van deliveries being replaceable by cargo bikes, due to their faster speeds, shorter parking times and more efficient routes across By modelling relative performance different vehicle types micro-regions, machine learning can help operators evaluate business...
We present PyTorch Geometric Temporal a deep learning framework combining state-of-the-art machine algorithms for neural spatiotemporal signal processing. The main goal of the library is to make temporal geometric available researchers and practitioners in unified easy-to-use framework. was created with foundations on existing libraries eco-system, streamlined network layer definitions, snapshot generators batching, integrated benchmark datasets. These features are illustrated tutorial-like...