Generating Financial Time Series: Benchmarks, Rough Path Methods, and Hands-On Evaluation
Organizers: Anthony K. H. Tung (National University of Singapore), Hao Ni (University College London), Yihao Ang (National University of Singapore), Yifan Bao (National University of Singapore), Xinyu Xi (National University of Singapore)
Description: This tutorial introduces the emerging field of synthetic financial time series generation, a powerful tool for overcoming data scarcity, ensuring privacy, and enabling robust model development in finance. Participants will gain a comprehensive understanding of the landscape of generative models, ranging from GANs, VAEs, and diffusion models to recent benchmark initiatives, and how they perform across fidelity, tradability, and robustness. A central focus will be rough-path-based methods (e.g., SigWGAN, PCF-GAN), which offer mathematically principled approaches to capturing complex financial dynamics and are increasingly adopted in industry. The tutorial combines foundational talks with an interactive hands-on session, where participants can evaluate whether synthetic data not only looks realistic but also supports profitable trading strategies under real-world frictions. By bridging theory, benchmarks, and practice, this tutorial equips researchers and practitioners with the tools to design, evaluate, and deploy financially meaningful generative models.
Website: https://sites.google.com/view/financial-tsg
Robust Graph Learning in Finance
Organizers: Xiang Ao (Institute of Computing Technology, Chinese Academy of Sciences), Yang Liu (Institute of Computing Technology, Chinese Academy of Sciences), Guansong Pang (Singapore Management University), Yuanhao Ding (Institute of Computing Technology, Chinese Academy of Sciences), Hezhe Qiao (Singapore Management University), Dawei Cheng (Tongji University), Qing He (Institute of Computing Technology, Chinese Academy of Sciences)
Description: Graph learning is vital for financial analytics, enabling tasks like fraud detection. However, most models assume ideal, stable data, which is unrealistic. Real-world financial systems face volatility, errors, and adversarial attacks that degrade model performance. This tutorial surveys strategies for robust graph learning in finance. It classifies specific robustness threats and categorizes current robust graph learning approaches, spanning data-level preprocessing to model-level adaptation and generalization, and discusses representative techniques in detail. Real-world case studies illustrate how these challenges manifest and are addressed in production. Integrating theory with practice, it provides actionable strategies to secure graph-based AI in high-stakes financial environments.
Website: https://qwer12191.github.io/robust-graph-learning/
Bridging Prediction and Optimization: Decision-Focused Learning in Financial Optimization
Organizers: Yongjae Lee (UNIST), Woo Chang Kim (KAIST), Junhyeong Lee (UNIST), Hyunglip Bae (KAIST), Haeun Jeon (KAIST)
Description: Decision-Focused Learning (DFL) is an emerging paradigm that directly aligns machine learning with the quality of downstream decisions. This tutorial provides the first comprehensive introduction to DFL in the context of financial optimization, bridging predictive modeling with decision-making under uncertainty. We will cover the foundations of DFL, its theoretical underpinnings, and its contrasts with prediction-focused learning (PFL). Through detailed case studies, we demonstrate how DFL reshapes return prediction in mean-variance portfolio optimization, enhances goal-based investing, and addresses partial index tracking with real-world constraints. The tutorial combines theory, live coding sessions, and hands-on exercises using PyTorch and differentiable optimization layers, equipping participants with both conceptual insights and practical skills. By the end, attendees will understand how DFL models trade off predictive accuracy to achieve superior financial decisions, and they will gain access to open-source code resources for further exploration.
Website: https://bridge-po.github.io/
AI in Financial Services: Risks and Opportunities for Compliant Decision Intelligence
Organizers: Garbriele Mazzini (MIT), Svitlana Vyetrenko (Outsampler, University of Strasbourg)
Description: Artificial Intelligence is transforming the financial services industry, from algorithmic trading and risk management to fraud detection and customer personalization. Yet, adoption is constrained by regulatory requirements, explainability concerns, and the operational complexity of deploying AI in high-stakes environments. This tutorial provides a guide to building, evaluating, and deploying AI systems in financial services, with an emphasis on interpretable, compliant, and robust models. Participants will
leave with a clear roadmap for developing AI solutions in financial services, practical tools for explainability, compliance, and deployment, and an awareness of common pitfalls in production AI systems and how to avoid them.
Website: https://sites.google.com/view/icaif-2025-ai-compliance
ACM ICAIF'25 