Paper: arXiv 2403.15243

Authors: Florian Krach, Josef Teichmann, Hanna Wutte

Abstract

Robust utility optimization enables an investor to deal with market uncertainty in a structured way, with the goal of maximizing the worst-case outcome. In this work, we propose a generative adversarial network (GAN) approach to (approximately) solve robust utility optimization problems in general and realistic settings. In particular, we model both the investor and the market by neural networks (NN) and train them in a mini-max zero-sum game. This approach is applicable for any continuous utility function and in realistic market settings with trading costs, where only observable information of the market can be used. A large empirical study shows the versatile usability of our method. Whenever an optimal reference strategy is available, our method performs on par with it and in the (many) settings without known optimal strategy, our method outperforms all other reference strategies. Moreover, we can conclude from our study that the trained path-dependent strategies do not outperform Markovian ones. Lastly, we uncover that our generative approach for learning optimal, (non-) robust investments under trading costs generates universally applicable alternatives to well known asymptotic strategies of idealized settings.

Complexity vs Empirical Score

  • Math Complexity: 8.0/10
  • Empirical Rigor: 7.5/10
  • Quadrant: Holy Grail — high math complexity, high empirical rigor

Why this score: The paper employs advanced mathematical concepts such as stochastic calculus, minimax optimization, and deep learning theory, representing high math complexity. It includes a large empirical study with comparative metrics against reference strategies, demonstrating substantial implementation and data-driven validation.

Research Flowchart

  flowchart TD
  A["Research Goal: Robust Utility Optimization via GAN Approach"] --> B["Methodology: GAN with Investor & Market NNs"]
  B --> C["Input: Realistic Market Data with Trading Costs"]
  C --> D["Process: Mini-Max Zero-Sum Game Training"]
  D --> E{"Outcomes"}
  E --> F["Performs on par with optimal strategies where available"]
  E --> G["Outperforms other strategies in unknown settings"]
  E --> H["Markovian strategies suffice over path-dependent ones"]
  E --> I["Generates universal alternatives to idealized asymptotic strategies"]

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 a generative adversarial network (GAN) approach to solve robust utility optimization problems in realistic market settings with trading costs.

Source sentence
In this work, we propose a generative adversarial network (GAN) approach to (approximately) solve robust utility optimization problems in general and realistic settings.

Assumes 5

  • Both the actor and the market networks only receive current stock prices and portfolio value as input, not each other's outputs. market
    Source sentence
    both the actor and the market network only get the current stock prices as well as the current portfolio value as input, but not the outputs of each other.
  • Utility functions are continuous and can be any shape (e.g., log, power). model
    Source sentence
    This approach is applicable for any continuous utility function
  • Path‑wise constraints are imposed on the adversarial market measures. model
    Source sentence
    We introduce fairly general, path-wise constraints for adversarial market measures
  • Training data consist of Gaussian increments. data
    Source sentence
    D {training set of gaussian increments}
Show 1 more
  • Early stopping based on validation metrics approximating worst‑case expected utility. other
    Source sentence
    we make use of this by early stopping criteria based on validation metrics approximating the worst-case expected utility

Data

  • Gaussian increments synthetic unknown
    Source sentence
    D {training set of gaussian increments}

Costs

Modeled Costs are part of the reported headline results.

Finds 5

  • GAN approaches successfully recover optimal strategies in multi‑asset friction‑less markets with log utilities. performance
    Source sentence
    We empirically show that the GAN approaches successfully recover the optimal strategies in multi-asset, friction-less markets with log utilities.
  • With trading costs, GAN‑based robust strategies outperform explicit friction‑less solutions. performance
    Source sentence
    we found that with trading costs, GAN-based robust strategies outperform explicit strategies that were obtained as solutions in the friction-less setting.
  • Non‑robustly trained GAN strategies exhibit robust behavior by reducing risky investments compared to asymptotically optimal strategies. performance
    Source sentence
    even non-robustly trained GAN-based strategies displayed robust behavior, meaning that they learned to overall reduce risky investments compared to asymptotically optimal investment strategies
  • Feed‑forward and recurrent neural network architectures yield comparable performance. performance
    Source sentence
    we saw comparable performance with feed-forward and recurrent NN architectures
  • Fully robust GAN strategy outperforms non‑robust GAN and pure bank‑account strategies in noisy market evaluations. performance
    Source sentence
    other strategies in the evaluation with noisy markets

Checked by others

No other paper in the archive has been found to support or contradict these findings (compared with 48 related papers). Coverage is partial, so this is neither replication nor refutation.

Extracted 2026-11-14 · same card as JSON · A language model chose these items: check the source sentence before relying on one.

Cited References (26)

  1. Robust risk-aware option hedging — → 2303.15216
  2. Robust Hedging GANs — → 2307.02310
  3. Robust Utility Optimization via a GAN Approach (2024) — link
  4. Building an Optimal Portfolio in Discrete Time in the Presence of Transaction Costs (2010)
  5. Robust risk-aware reinforcement learning (2022)
  6. Robust Utility Maximization in Discrete-Time Markets with Friction (2018)
  7. Sequential Decision Making With Coherent Risk (2017) — link
  8. Robust utility maximization under model uncertainty via a penalization approach (2022)
and 18 more
  1. Adaptive Robust Control Under Model Uncertainty (2017) — link
  2. Generative adversarial networks (2020)
  3. Unsupervised Representation Learning with Deep Convolutional Generative Adversarial Networks (2016) — link
  4. Recent advances and application of generative adversarial networks in drug discovery, development, and targeting (2022)
  5. Generative Adversarial Networks for Image and Video Synthesis: Algorithms and Applications (2020) — link
  6. Tail- GAN : Learning to Simulate Tail Risk Scenarios (2023) — link
  7. Stochastic portfolio theory: An Overview (2009)
  8. A Primer on Portfolio Choice with Small Transaction Costs (2017) — link
  9. Lifetime portfolio selection under uncertainty: The continuous-time case (1969)
  10. Optimal investment under transaction costs: A threshold rebalanced portfolio approach (2013)
  11. Optimal investment for retail investors (2022)
  12. Portfolio selection with transaction costs (1990)
  13. Optimal trading with linear costs (2012) — link
  14. Multiperiod consumption and investment behavior with convex transactions costs (1979)
  15. Optimal multifactor trading under proportional transaction costs (2012) — link
  16. Universal trading under proportional transaction costs (2016) — link
  17. Learning of the No-Trade Zone (2019)
  18. Portfolio choice with return predictability and small trading frictions (2022)

2 of 26 cited works are in the corpus; resolved via arXiv source bundles, Semantic Scholar and OpenAlex. Corpus-wide: 2596 linked papers.