Paper: arXiv 2509.06702
Authors: Ruben Bontorno, Songyan Hou
Abstract
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.
Complexity vs Empirical Score
- Math Complexity: 8.0/10
- Empirical Rigor: 7.0/10
- Quadrant: Holy Grail — high math complexity, high empirical rigor
Why this score: The paper employs advanced optimal transport theory and stochastic calculus with dense mathematical notation, including definitions, propositions, and convergence theorems. It includes rigorous empirical validation with code-ready algorithms, statistical convergence tests, and benchmarking against theoretical values on synthetic financial processes.
Research Flowchart
flowchart TD A["Research Goal<br>Evaluate generative models for financial time series<br>in decision-making tasks"] --> B["Key Methodology<br>Nested Optimal Transport Distance<br>time-causal & robust to hedging/RL"] B --> C["Data Inputs<br>Financial time series datasets<br>e.g., stock prices"] C --> D["Computational Process<br>Statistically consistent, parallelizable algorithm<br>for efficient computation"] D --> E["Key Findings & Outcomes<br>Substantial speedups over existing approaches<br>Validated for hedging, optimal stopping, RL"]
Evidence Card
What the authors claim, on what data, and who has checked it. Read from the paper's own text by a language model: every quote is verbatim, nothing was re-run. How to read this card
Proposes
The paper proposes the nested optimal transport (AW) distance as a robust metric for evaluating generative AI on financial time series and introduces a naturally parallelizable algorithm to compute it efficiently.
Source sentence
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
Assumes 2
- Adapted empirical measures converge to the underlying distribution as the number of samples N goes to infinity. statistical
Source sentence
For AW-distances, empirical measures fail to converge, but adapted empirical measures converge to the underlying measure as the number of samples N goes to infinity, which provides statistical guarantee approximating µ with ˆµN in calculating AW-distances; see the proof of Theorem 2 in Appendix A.3. Theorem 2. Let µ ∈P2(RdT ) and ∆N = N −1 dT .
- When the underlying process is Markovian, a modified adapted empirical measure yields faster convergence rates independent of the time horizon T. model
Source sentence
However, if µ is known to be Markovian, there is a natural Markovian modification of the adapted empirical measure, simply replacing (ˆµN)t+1 q1:t by (ˆµN)t+1 qt , which can be similarly defined as (ˆµN)t+1 q1:t ; see [14]. This modification yields substantially improved statistical convergence rates O(N −1 2d ) that are independent of T; see Theorem 6.1 in [14].
Data
- Ornstein-Uhlenbeck processes synthetic unknown
Source sentence
Ornstein-Uhlenbeck processes. Let us first consider the Ornstein-Uhlenbeck (OU) processes Xσ = (Xσ t )t≥0, defined by dXσ t = −Xσ t dt + σdWt, X0 = 0, where (Wt)t≥0 is a Brownian motion and σ ∈R+.
- Fake Brownian Motion synthetic unknown
Source sentence
Fake Brownian Motion. Let (Wt)t∈[0,1] be a Brownian motion, Z ∼N(0, 1) independent of (Wt)t∈[0,1] and δ ∈(0, 1).
Costs
Not applicable
Finds 3
- The proposed parallel algorithm achieves orders-of-magnitude speedups over existing implementations for computing AW distances. performance
Source sentence
Moreover, we propose a statistically consistent, naturally parallelizable algorithm for its computation, achieving substantial speedups over existing approaches. In this paper, we propose a natural parallelization algorithm to calculate AW-distances, which achieves orders-of-magnitude speedups over existing implementations and also behaves statistically consistently.
- Numerical values of AW2^2 computed with the Markovian implementation converge to the theoretical values for Ornstein-Uhlenbeck processes. statistical
Source sentence
Denoting by µσ the distribution of ˜Xσ 1:N, we compute AW2 2(µ1, µ3) with randomly generated i.i.d. samples using the Markovian implementation of our algorithm, and evidently, numerical values of AW2 2 converge to its theoretical value; see Figure 1.
- The Markovian modification of adapted empirical measures yields substantially improved statistical convergence rates O(N^{-1/2d}) independent of the time horizon. theoretical
Source sentence
This modification yields substantially improved statistical convergence rates O(N −1 2d ) that are independent of T; see Theorem 6.1 in [14].
Checked by others
No other paper in the archive has been found to support or contradict these findings (compared with 15 related papers). Coverage is partial, so this is neither replication nor refutation.
Code
No repository by the authors found in the paper.
Extracted 2026-11-14 · same card as JSON · A language model chose these items: check the source sentence before relying on one.
Cited References (87)
- From empirical observations to tree models for stochastic optimization: convergence properties (2016)
- Causal transference plans and their Monge-Kantorovich problems (2018)
- Causal transport in discrete time and applications (2017)
- Causal optimal transport and its links to enlargement of filtrations and continuous-time stochastic optimization (2020)
- Changes of filtrations and of probability measures (1978)
- Adapted Wasserstein distances and stability in mathematical finance (2020)
- The Wasserstein distance and approximation theorems (1985)
- A distance for multistage stochastic optimization models (2012)
and 79 more
- Multistage stochastic optimization (2014)
- Dynamic generation of scenario trees (2015)
- Incorporating statistical model error into the calculation of acceptability prices of contingent claims (2019)
- Evaluations of risk measures for different probability measures (2013)
- On the geometry of the space of probability measures in Rn endowed with the quadratic optimal transport distance (2008)
- On a Wasserstein-type distance between solutions to stochastic differential equations (2019)
- Convergence of smoothed empirical measures with applications to entropy estimation (2020)
- Asymptotic guarantees for generative modeling based on the smooth wasserstein distance (2020)
- Convergence of Gaussian-smoothed optimal transport distance with sub-gamma distributions and dependent samples (2021)
- Limit Distribution Theory for the Smooth 1-Wasserstein Distance with Applications (2021)
- Limit distribution theory for smooth $p$-Wasserstein distances (2022)
- On the method of bounded differences (1989)
- Gaussian-Smoothed Optimal Transport: Metric Structure and Statistical Efficiency (2020)
- Rate of convergence of the smoothed empirical Wasserstein distance (2022)
- Smooth p-Wasserstein Distance: Structure, Empirical Approximation, and Statistical Applications (2021)
- Asymptotics of Smoothed Wasserstein Distances (2020)
- Nonparametric mixture MLEs under Gaussian-smoothed optimal transport distance (2021)
- Statistical inference with regularized optimal transport (2022)
- An extension of Tietze’s theorem. (1951)
- Estimating processes in adapted Wasserstein distance (2022)
- Sensitivity of Multiperiod Optimization Problems with Respect to the Adapted Wasserstein Distance (2023)
- All adapted topologies are equal (2019)
- A note on the adapted weak topology in discrete time (2022) — link
- The Wasserstein space of stochastic processes (2024)
- Quantitative fundamental theorem of asset pricing (2022) — link
- Mathematical foundations of infinite-dimensional statistical models (2021)
- On the mean speed of convergence of empirical and occupation measures in Wasserstein distance (2011)
- The speed of mean Glivenko-Cantelli convergence (1969)
- Quantitative concentration inequalities for empirical measures on non-compact spaces (2007)
- On the mean speed of convergence of empirical and occupation measures in Wasserstein distance (2014)
- Constructive quantization: Approximation by empirical measures (2013)
- Convergence and concentration of empirical measures under Wasserstein distance in unbounded functional spaces (2018)
- Sharp asymptotic and finite-sample rates of convergence of empirical measures in Wasserstein distance (2019)
- On the rate of convergence in Wasserstein distance of the empirical measure (2015)
- Concentration of empirical distribution functions with applications to non-i.i.d. models (2010)
- Convergence of the empirical measure in expected Wasserstein distance: non asymptotic explicit bounds in Rd (2022)
- Entropic optimal transport between unbalanced gaussian measures has a closed form (2020)
- The statistical effect of entropic regularization in optimal transportation (2020) — link
- Adapted optimal transport between Gaussian processes in discrete time (2025)
- Entropy-regularized 2-Wasserstein distance between Gaussian measures (2022)
- The schr "o dinger bridge between gaussian measures has a closed form (2023)
- Computational methods for adapted optimal transport (2024)
- The nested Sinkhorn divergence to learn the nested distance (2022)
- Stability of entropic optimal transport and Schr "o dinger bridges (2022)
- Quantitative Stability of Regularized Optimal Transport and Convergence of Sinkhorn’s Algorithm (2022)
- Quantitative uniform stability of the iterative proportional fitting procedure (2021) — link
- A differential approach to the multi-marginal Schrödinger system (2020)
- Matching in closed-form: equilibrium, identification, and comparative statics (2016)
- Convergence of the adapted smoothed empirical measures (2024) — link
- Deep generative learning via schr "o dinger bridge (2021)
- Wasserstein upper bounds of the total variation for smooth densities (2020)
- Compactness in adapted weak topologies (2019) — link
- Separate continuity and measurability (1969)
- Bounding adapted Wasserstein metrics (2024) — link
- Entropic adapted Wasserstein distance on Gaussians (2024) — link
- Fitted value iteration methods for bicausal optimal transport — → 2306.12658
- Measure estimation on manifolds: an optimal transport approach (2022)
- Weakly differentiable functions: Sobolev spaces and functions of bounded variation (2012)
- Time consistent dynamic risk processes (2009)
- Sensitivity analysis of Wasserstein distributionally robust optimization problems (2021)
- Minimax estimation of smooth densities in Wasserstein distance (2022)
- Sensitivity of multiperiod optimization problems with respect to the adapted Wasserstein distance (2023)
- On choosing and bounding probability metrics (2002)
- Lipschitz continuity of probability kernels in the optimal transport framework (2023)
- On uniform continuity of posterior distributions (2020)
- Wasserstein upper bounds of Lp-norms for multivariate densities in Besov spaces (2024)
- Convergence of adapted empirical measures on $\mathbb R ^d$ (2024)
- Comparison between $W_2$ distance and $\dot H ^ -1 $ norm, and localization of Wasserstein distance (2018)
- Approximate earth mover’s distance in linear time (2008)
- Causal transport on path space (2024) — link
- Bicausal optimal transport for SDEs with irregular coefficients (2024) — link
- The fast rate of convergence of the smooth adapted Wasserstein distance (2025) — link
- Time-Causal VAE: Robust Financial Time Series Generator — → 2411.02947
- Generating synthetic data in finance: opportunities, challenges and pitfalls (2020)
- Deep generative modeling for financial time series with application in VaR: A comparative review — → 2401.10370
- Synthetic Data for Portfolios: A Throw of the Dice Will Never Abolish Chance — → 2501.03993
- Generative model for financial time series trained with MMD using a signature kernel — → 2407.19848
- On Wasserstein geometry of Gaussian measures (2010)
- Pot: Python optimal transport (2021)
5 of 87 cited works are in the corpus; resolved via arXiv source bundles, Semantic Scholar and OpenAlex. Corpus-wide: 2596 linked papers.