My name is Gengchen Mai, a Ph.D. candidate at Space and Time for Knowledge Organization Lab, Department of Geography, University of California, Santa Barbara. My Ph.D. adviser is Prof. Krzysztof Janowicz. I am interested in Machine Learning/Deep Learning, Geographical Information Science (GIScience), Geographic Question Answering, NLP, Geographic Information Retrieval, Knowledge Graph, and Semantic Web. Right now, my research is highly focused on Geographic Question Answering and Spatially-Explicit Machine Learning models. I have completed four AI/ML research-based internships at Esri Inc., SayMosaic Inc., Apple Map, and Google X.
My research mainly focuses on combining top-down, theory-informed methodologies, and bottom-up, data-driven approaches for geospatial knowledge discovery and tackling global challenges. It covers a wide range of topics including GeoAI, Spatial Data Mining, Geospatial Knowledge Graph, Geographic Knowledge Graph, and Geographic Information Retrieval. The first part of my Ph.D. thesis work on Geographic Question Answering over Knowledge Graph has won the 1st place best full paper award at AGILE 2019, one of the top conferences at GIScience. Currently, I am particularly interested in developing spatially-explicit machine learning models that aim at using spatial principles and spatial inductive bias to improve the performance of the state-of-the-art ML models on different domains such as Urban Science, Ecology, and Earth Science. My recent publication on this topic about a multi-scale representation learning model for spatial feature distributions has been accepted as a spotlight paper at ICLR 2020 which has been widely recognized as one of the top conferences in the AI/ML field (ranked 17 in Google Scholar 2020 Metrics) with a very low acceptance rate (6%). This research has been selected as one of the 16 best deep learning papers of ICLR 2020 by Neptune.ai. Several following works of mine have shown the effectiveness of this idea on multiple tasks across different domains such as geographic question answering (NLP), geo-aware image classification (CV), POI representation learning (Urban Data Science).
1.Gengchen Mai, Krzysztof Janowicz, Ling Cai, Rui Zhu, Blake Regalia, Bo Yan, Meilin Shi, Ni Lao. SE-KGE: A Location-Aware Knowledge Graph Embedding Model for Geographic Question Answering and SpatialSemantic Lifting.Transactions in GIS.DOI:10.1111/TGIS.12629
2. Gengchen Mai, Krzysztof Janowicz, Sathya Prasad, Meilin Shi, Ling Cai, Rui Zhu, Blake Regalia, Ni Lao.Semantically-Enriched Search Engine for Geoportals: A Case Study with ArcGIS Online, In: Proceedings of AGILE 2020, Jun. 16 - 19, 2020, Chania, Crete, Greece.(Acceptance Rateâ‰ˆ35%)
3.Gengchen Mai, Krzysztof Janowicz, Bo Yan, Rui Zhu, Ling Cai, Ni Lao. Multi-Scale Representation Learning for Spatial Feature Distributions using Grid Cells, In: Proceedings of ICLR 2020, Apr. 26 - 30, 2020, Addis Ababa, ETHIOPIA.* Spotlight Paper (Acceptance Rate 6%, 156 out of 2594 submissions)
4.Gengchen Mai, Krzysztof Janowicz, Bo Yan, Rui Zhu, Ling Cai, Ni Lao. Contextual Graph Attention for Answering Logical Queries over Incomplete Knowledge Graphs, In: Proceedings of ACM K-CAP 2019, Nov. 19- 21, 2019, Marina del Rey, CA, USA.(Acceptance Rate 25%)
5. Ling Cai, Bo Yan, Gengchen Mai, Krzysztof Janowicz, Rui Zhu. TransGCN: A Translation-Based GraphConvolutional Network Model for Link Prediction, In: Proceedings of ACM K-CAP 2019, Nov. 19 - 21, 2019, Marina del Rey, CA, USA.* 1st Best Full Paper Award (1 out of 28 accepted full papers)
6.Gengchen Mai, Bo Yan, Krzysztof Janowicz, Rui Zhu. Relaxing Unanswerable Geographic Questions UsingA Spatially Explicit Knowledge Graph Embedding Model, In: Proceedings of AGILE 2019, June 17 - 20, 2019, Limassol, Cyprus.* 1st Best Full Paper Award (1 out of 19 accepted full papers)
7.Gengchen Mai, Krzysztof Janowicz, Bo Yan. Support and Centrality: Learning Weights for Translation-based Knowledge Graph Embedding Models, In: Proceedings of EKAW 2018, Nov. 12 - 16, 2018, Nancy, France.(Acceptance Rate 26%)
8.Gengchen Mai, Krzysztof Janowicz, Bo Yan, Simon Scheider. Deeply Integrating Linked Data with Geo-graphic Information Systems. Transactions in GIS.DOI:10.1111/tgis.12538
9. Bo Yan, Krzysztof Janowicz, Gengchen Mai, Rui Zhu. A Spatially-Explicit Reinforcement Learning Model for Geographic Knowledge Graph Summarization.Transactions in GIS.DOI:10.1111/tgis.12547
10.Gengchen Mai, Krzysztof Janowicz, Yingjie Hu, Song Gao. ADCN: An Anisotropic Density-Based ClusteringAlgorithm for Discovering Spatial Point Patterns with Noise.Transactions in GIS, 22 (2018) 348-369.DOI:10.1111/tgis.12313* Top 10% Most Downloaded Papers in TGIS (01/2018-12/2019)