Welcome to DeepSpatial 2024

Title: Is deep learning fit for purpose for time series classification tasks?

Abstract: The recent “Bake-off redux” benchmarking exercise that compared a large range of state-of-the-art time series classification algorithms across a large benchmark of 112 tasks found that no deep learning methods were overall competitive with the state of the art. I discuss this finding from the perspective of an active deep learning for time series classification researcher and address the question of whether deep learning for time series classification has a future.

Bio: Professor Geoff Webb of the Monash University Department of Data Science and Artificial Intelligence is an eminent and highly-cited data scientist. He was editor in chief of the Data Mining and Knowledge Discovery journal, from 2005 to 2014. He has been Program Committee Chair of both ACM SIGKDD and IEEE ICDM, as well as General Chair of ICDM and member of the ACM SIGKDD Executive. He is a Technical Advisor to machine learning as a service startup BigML Inc and to recommender systems startup FROOMLE. He has developed many useful machine learning algorithms that are widely deployed, including many of the current leading approaches to time series classification. His many awards include IEEE Fellow, the inaugural Eureka Prize for Excellence in Data Science (2017), the Pacific-Asia Conference on Knowledge Discovery and Data Mining Distinguished Research Contributions Award (2022) and the IEEE International Conference on Data Mining 10-year Highest Impact Award (2023). He has thrice been recognised by The Australian Research Magazine as Australia’s leading Bioinformatics and Computational Biology researcher.

Title: Towards A Universal Model for Urban Spatio-Temporal Prediction: A Knowledge-informed Generative Learning Perspective.

Abstract: Urban spatio-temporal prediction is essential for effective decision-making in areas such as transportation management and resource optimization. While cutting-edge prediction models have advanced, they often address specific problems without generalizing across different domains. Generative models, particularly Generative Pre-trained Transformers (GPT) and Diffusion Probabilistic Models, have demonstrated significant capabilities in text and video applications. However, their potential for urban spatio-temporal prediction remains largely unexplored. This talk will introduce a versatile generative learning framework that leverages urban spatio-temporal knowledge for universal modeling. The presentation will cover 1) the construction of an urban knowledge graph to capture semantic relationships among spatial regions, temporal context, and urban flows; 2) the design of physics-informed diffusion models to characterize intrinsic mechanisms behind spatio-temporal data; and 3) the development of knowledge-guided prompt learning to create a one-for-all solution applicable across diverse scenarios.

Bio: Dr. Li is currently a Professor of the Department of Electronic Engineering, Tsinghua University. He received the Ph.D. degree in electronic engineering from Tsinghua University in 2012. His research interests include city science and urban computing. Dr. Li has served as General Chair, TPC Chair, SPC/TPC Member for several international workshops and conferences, and he is on the editorial board of two IEEE journals. He has published over 100 papers on first-tier international conferences and journals, including Nature Machine Intelligence, Nature Human Behavior, Nature Computational Science, KDD, NeurIPS, WWW, UbiComp, and his papers have total citations more than 25000. Among them, five receive conference Best Paper (run-up) Awards.
Jingtao Ding
Dr. Ding is currently a Postdoctoral Research Fellow in the Department of Electronic Engineering at Tsinghua University. He obtained his Ph.D. Degree and Bachelor’s Degree from the same department in 2020 and 2015, respectively. During the years 2020 and 2022, he worked as an applied researcher at Wechat, the largest SNS company in China. His current research focuses primarily on generative modeling of complex human behavior and dynamics in urban system. He has over 40 publications in top-tier venues on data science and machine learning, such as KDD, NeurIPS, WWW, ICLR, IJCAI, IEEE TKDE, ACM TOIS, etc. He serves as the academic editor for PLOS Complex Systems and PC member for several top-tier conferences. His work on understanding user behaviors in the complex network of recommender system won the Best Poster Paper Award in WWW 2018.

Title: Data-Efficient Learning for Imperfect Spatio-Temporal Data: Leveraging SSLs and LLMs

Abstract: Spatio-temporal data have been very instrumental in developing predictive models in different domains, including for energy, transport and mobility, retail, and public health management. However, modelling with spatiotemporal data is hampered by the limited availability and sparsity of training data, due to the lack of labels and/or semantic information on spatio-temporal data, and issues with the data, including noise and missing data. This talk explores recent advancements in self-supervised learning (SSL) and neural architectures for spatio-temporal data. SSL, particularly contrastive learning, has shown promise in learning discriminative representations without labeled data, yet existing methods often struggle with multimodal data integration and computational inefficiencies. To address these challenges, we introduce several innovative models: a cross-modal SSL approach that can deal with missing sensor data, and a neural-ODE architecture that can handle sporadic time-series data. To address data sparsity and lack of semantic information in spatio-temporal data, we leverage LLMs to improve the generalizability, in zero-shot and fine-tuned settings. Examples will include the use of LLMs in dynamic contextual information embedding, mobility forecasting, next point-of-interest (POI) recommendation task, and energy load forecasting.

Bio: Flora Salim is a Professor in the School of Computer Science and Engineering (CSE) and the inaugural Cisco Chair of Digital Transport & AI, University of New South Wales (UNSW) Sydney, and the Deputy Director (Engagement) of UNSW AI Institute. Her research is on generalisable and data-efficient machine learning for multimodal spatio-temporal learning tasks. She is a member of the Australian Research Council (ARC) College of Experts. She currently serves as an Editor of IMWUT, Associate Editor of ACM TSAS, and an Associate Editor-in-Chief of IEEE Pervasive Computing, and an Area Chair for many conferences including NeurIPS, ICLR, and The Web Conf (WWW). She has worked with many industry and government partners, and managed large-scale research and innovation projects, leading to several patents and deployed systems.

Title: TBA

Abstract: TBA

Bio: TBA