Schedule

Abstract: In this presentation, Dr. Zhang will provide an overview of AI research on foundational models for time series data. The talk will include an industrial perspective on time series research and its applications in facilitating real-world business solutions. Dr. Zhang will also explore Ant Group's innovative work in this area, focusing on advancements such as TimeLLM, TimeMixer, and iTransformer, along with the philosophical considerations underpinning these developments. The presentation will provide a discussion on real-world applications and the potential implications for future academic and industrial endeavors in this field.

We show language models that have only been trained for next word prediction can outperform purpose-built time-series models trained on time series training data. How is this result possible? What does it mean for the future of time series forecasting? In this talk, I'll tell the story of how we proposed the first large language model for time series forecasting, how this field has since been evolving, and explain why this class of models has promise in domains outside of language.

Speaker: Yifu Cai. Title: Do LLMs understand financial series?

Speaker: Andrew Robert Williams and Arjun Ashok. Title: Context is Key: A benchmark for forecasting with essential textual information.

Speaker: Seonkyu Lim. Title: Enhanced Fraud Detection in Bank Transfers via Augmented Account Features.