Nested Optimal Transport Distances
This work addresses the problem of evaluating generative AI for financial time series in decision-making applications, offering a practical solution for stress testing and scenario generation.
The paper tackled the lack of consensus metrics for evaluating deep generative models of financial time series by employing the nested optimal transport distance, a robust variant for tasks like hedging and reinforcement learning, and proposed a statistically consistent algorithm that achieves substantial speedups over existing approaches.
Simulating realistic financial time series is essential for stress testing, scenario generation, and decision-making under uncertainty. Despite advances in deep generative models, there is no consensus metric for their evaluation. We focus on generative AI for financial time series in decision-making applications and employ the nested optimal transport distance, a time-causal variant of optimal transport distance, which is robust to tasks such as hedging, optimal stopping, and reinforcement learning. Moreover, we propose a statistically consistent, naturally parallelizable algorithm for its computation, achieving substantial speedups over existing approaches.