IDEAS home Printed from https://ideas.repec.org/p/arx/papers/2603.06563.html

Convergence of Neural Network Policies for Risk--Reward Optimization

Author

Listed:
  • Chang Chen
  • Duy-Minh Dang

Abstract

We develop a neural-network framework for multi-period risk--reward stochastic control problems with constrained two-step feedback policies that may be discontinuous in the state. We allow a broad class of objectives built on a finite-dimensional performance vector, including terminal and path-dependent statistics, with risk functionals admitting auxiliary-variable optimization representations (e.g.\ Conditional Value-at-Risk and buffered probability of exceedance) and optional moment dependence. Our approach parametrizes the two-step policy using two coupled feedforward networks with constraint-enforcing output layers, reducing the constrained control problem to unconstrained training over network parameters. Under mild regularity conditions, we prove that the empirical optimum of the NN-parametrized objective converges in probability to the true optimal value as network capacity and training sample size increase. The proof is modular, separating policy approximation, propagation through the controlled recursion, and preservation under the scalarized risk--reward objective. Numerical experiments confirm the predicted convergence-in-probability behavior, show close agreement between learned and reference control heat maps, and demonstrate out-of-sample robustness on a large independent scenario set.

Suggested Citation

  • Chang Chen & Duy-Minh Dang, 2026. "Convergence of Neural Network Policies for Risk--Reward Optimization," Papers 2603.06563, arXiv.org.
  • Handle: RePEc:arx:papers:2603.06563
    as

    Download full text from publisher

    File URL: http://arxiv.org/pdf/2603.06563
    File Function: Latest version
    Download Restriction: no
    ---><---

    References listed on IDEAS

    as
    1. S. G. Kou, 2002. "A Jump-Diffusion Model for Option Pricing," Management Science, INFORMS, vol. 48(8), pages 1086-1101, August.
    2. Anders Max Reppen & Halil Mete Soner, 2023. "Deep empirical risk minimization in finance: Looking into the future," Mathematical Finance, Wiley Blackwell, vol. 33(1), pages 116-145, January.
    3. Dang, D.M. & Forsyth, P.A., 2016. "Better than pre-commitment mean-variance portfolio allocation strategies: A semi-self-financing Hamilton–Jacobi–Bellman equation approach," European Journal of Operational Research, Elsevier, vol. 250(3), pages 827-841.
    4. Li, Yuying & Forsyth, Peter A., 2019. "A data-driven neural network approach to optimal asset allocation for target based defined contribution pension plans," Insurance: Mathematics and Economics, Elsevier, vol. 86(C), pages 189-204.
    5. van Staden, Pieter M. & Dang, Duy-Minh & Forsyth, Peter A., 2021. "The surprising robustness of dynamic Mean-Variance portfolio optimization to model misspecification errors," European Journal of Operational Research, Elsevier, vol. 289(2), pages 774-792.
    6. Duy-Minh Dang & Chang Chen, 2025. "Multi-period Mean-Buffered Probability of Exceedance in Defined Contribution Portfolio Optimization," Papers 2505.22121, arXiv.org, revised Feb 2026.
    7. Pieter M. van Staden & Peter A. Forsyth & Yuying Li, 2024. "A Global-in-Time Neural Network Approach to Dynamic Portfolio Optimization," Applied Mathematical Finance, Taylor & Francis Journals, vol. 31(3), pages 131-163, May.
    8. A. Max Reppen & H. Mete Soner & Valentin Tissot-Daguette, 2023. "Deep stochastic optimization in finance," Digital Finance, Springer, vol. 5(1), pages 91-111, March.
    Full references (including those not matched with items on IDEAS)

    Most related items

    These are the items that most often cite the same works as this one and are cited by the same works as this one.
    1. German Nova Orozco & Duy-Minh Dang & Peter A. Forsyth, 2026. "Money-Back Tontines for Retirement Decumulation: Neural-Network Optimization under Systematic Longevity Risk," Papers 2602.16212, arXiv.org.
    2. Pieter van Staden & Peter Forsyth & Yuying Li, 2024. "Smart leverage? Rethinking the role of Leveraged Exchange Traded Funds in constructing portfolios to beat a benchmark," Papers 2412.05431, arXiv.org, revised Mar 2025.
    3. Hanwen Zhang & Duy-Minh Dang, 2023. "A monotone numerical integration method for mean-variance portfolio optimization under jump-diffusion models," Papers 2309.05977, arXiv.org.
    4. Zhang, Hanwen & Dang, Duy-Minh, 2024. "A monotone numerical integration method for mean–variance portfolio optimization under jump-diffusion models," Mathematics and Computers in Simulation (MATCOM), Elsevier, vol. 219(C), pages 112-140.
    5. Ni, Chendi & Li, Yuying & Forsyth, Peter, 2025. "Optimal multi-period leverage-constrained portfolios: A neural network approach," Journal of Economic Dynamics and Control, Elsevier, vol. 177(C).
    6. Chendi Ni & Yuying Li & Peter A. Forsyth, 2023. "Neural Network Approach to Portfolio Optimization with Leverage Constraints:a Case Study on High Inflation Investment," Papers 2304.05297, arXiv.org, revised May 2023.
    7. Peter Forsyth & Pieter van Staden & Yuying Li, 2025. "Making Leveraged Exchange-Traded Funds Work for your Portfolio," Papers 2506.19200, arXiv.org.
    8. Duy-Minh Dang & Hao Zhou, 2026. "Monotone 2D Integration Scheme for Mean-CVaR Optimization via Fourier-Trained Transition Kernels," Papers 2603.26291, arXiv.org.
    9. Peter A. Forsyth & Kenneth R. Vetzal & G. Westmacott, 2022. "Optimal performance of a tontine overlay subject to withdrawal constraints," Papers 2211.10509, arXiv.org.
    10. Duy-Minh Dang & Chang Chen, 2025. "Multi-period Mean-Buffered Probability of Exceedance in Defined Contribution Portfolio Optimization," Papers 2505.22121, arXiv.org, revised Feb 2026.
    11. van Staden, Pieter M. & Dang, Duy-Minh & Forsyth, Peter A., 2021. "The surprising robustness of dynamic Mean-Variance portfolio optimization to model misspecification errors," European Journal of Operational Research, Elsevier, vol. 289(2), pages 774-792.
    12. Forsyth, Peter A., 2020. "Optimal dynamic asset allocation for DC plan accumulation/decumulation: Ambition-CVAR," Insurance: Mathematics and Economics, Elsevier, vol. 93(C), pages 230-245.
    13. Chendi Ni & Yuying Li & Peter Forsyth & Ray Carroll, 2020. "Optimal Asset Allocation For Outperforming A Stochastic Benchmark Target," Papers 2006.15384, arXiv.org.
    14. Peter A. Forsyth & Yuying Li, 2025. "Risk Measures for DC Pension Plan Decumulation," Papers 2502.16364, arXiv.org.
    15. Pieter M. van Staden & Peter A. Forsyth & Yuying Li, 2023. "A parsimonious neural network approach to solve portfolio optimization problems without using dynamic programming," Papers 2303.08968, arXiv.org.
    16. Rong Du & Duy-Minh Dang, 2023. "Fourier Neural Network Approximation of Transition Densities in Finance," Papers 2309.03966, arXiv.org, revised Sep 2024.
    17. Jeonggyu Huh & Jaegi Jeon, 2024. "Pontryagin-Guided Policy Optimization for Merton's Portfolio Problem," Papers 2412.13101, arXiv.org, revised Jan 2025.
    18. Arash Fahim & Md. Arafatur Rahman, 2025. "Gaining efficiency in deep policy gradient method for continuous-time optimal control problems," Papers 2502.14141, arXiv.org, revised Feb 2025.
    19. van Staden, Pieter M. & Forsyth, Peter A. & Li, Yuying, 2024. "Across-time risk-aware strategies for outperforming a benchmark," European Journal of Operational Research, Elsevier, vol. 313(2), pages 776-800.
    20. Jeonggyu Huh & Seungwon Jeong & Hyun-Gyoon Kim & Hyeng Keun Koo & Byung Hwa Lim, 2026. "MarketGANs: Multivariate financial time-series data augmentation using generative adversarial networks," Papers 2601.17773, arXiv.org.

    More about this item

    NEP fields

    This paper has been announced in the following NEP Reports:

    Statistics

    Access and download statistics

    Corrections

    All material on this site has been provided by the respective publishers and authors. You can help correct errors and omissions. When requesting a correction, please mention this item's handle: RePEc:arx:papers:2603.06563. See general information about how to correct material in RePEc.

    If you have authored this item and are not yet registered with RePEc, we encourage you to do it here. This allows to link your profile to this item. It also allows you to accept potential citations to this item that we are uncertain about.

    If CitEc recognized a bibliographic reference but did not link an item in RePEc to it, you can help with this form .

    If you know of missing items citing this one, you can help us creating those links by adding the relevant references in the same way as above, for each refering item. If you are a registered author of this item, you may also want to check the "citations" tab in your RePEc Author Service profile, as there may be some citations waiting for confirmation.

    For technical questions regarding this item, or to correct its authors, title, abstract, bibliographic or download information, contact: arXiv administrators (email available below). General contact details of provider: http://arxiv.org/ .

    Please note that corrections may take a couple of weeks to filter through the various RePEc services.

    IDEAS is a RePEc service. RePEc uses bibliographic data supplied by the respective publishers.