Zhiyuan (Leo) Zhao
S1371CC
756 W Peachtree St NW
Atlanta, GA 30308
I’m a third-year Machine Learning Ph.D. student in Department of Computational Science and Engineering at Georgia Institute of Technology. I am affiliated with AdityaLab and advised by Dr. B Aditya Prakash. My research focuses on time-series forecasting with various techniques, such as attention, causality, physics-regularization, and LLMs. These research results have made real-world influences such as Flu and COVID-19 forecasting challenges organized by CDC.
I am also interested in pre-trained foundation models and applications of LLMs in time series tasks. Previously, my research spanned physics-informed neural networks with the advice of Prof. Aarti Singh and federated learning under the advice of Prof. Gauri Joshi at Carnegie Mellon University.
My current research focus includes:
- Out-of-distribution (OOD) generalization for time series;
- Time series fundation models and multi-modality time series forecasting;
- Time series forecasting in industrial application (with Amazon SCOT).
Selected Publications
- Time-MMD: A New Multi-Domain Multimodal Dataset for Time Series AnalysisNeural Information Processing Systems Datasets and Benchmarks Track, 2024
- Time-Series Forecasting for Out-of-Distribution Generalization Using Invariant LearningInternational Conference on Machine Learning, 2024
- LSTPrompt: Large Language Models as Zero-Shot Time Series Forecasters by Long-Short-Term PromptingFindings of the Association for Computational Linguistics, 2024
- Performative Time-Series ForecastingarXiv preprint arXiv:2310.06077, 2023
- PINNsFormer: A Transformer-Based Framework For Physics-Informed Neural NetworksInternational Conference on Learning Representations, 2024
- A dynamic reweighting strategy for fair federated learningIn ICASSP 2022-2022 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), 2022