Paper: arXiv 2504.04300

Authors: Anastasis Kratsios, Xiaofei Shi, Qiang Sun, Zhanhao Zhang

Abstract

We present a general computational framework for solving continuous-time financial market equilibria under minimal modeling assumptions while incorporating realistic financial frictions, such as trading costs, and supporting multiple interacting agents. Inspired by generative adversarial networks (GANs), our approach employs a novel generative deep reinforcement learning framework with a decoupling feedback system embedded in the adversarial training loop, which we term as the \emph{reinforcement link}. This architecture stabilizes the training dynamics by incorporating feedback from the discriminator. Our theoretically guided feedback mechanism enables the decoupling of the equilibrium system, overcoming challenges that hinder conventional numerical algorithms. Experimentally, our algorithm not only learns but also provides testable predictions on how asset returns and volatilities emerge from the endogenous trading behavior of market participants, where traditional analytical methods fall short. The design of our model is further supported by an approximation guarantee.

Complexity vs Empirical Score

  • Math Complexity: 8.5/10
  • Empirical Rigor: 4.5/10
  • Quadrant: Lab Rats — theoretically deep, empirically untested

Why this score: The paper provides a strong theoretical and mathematical foundation, combining deep reinforcement learning, generative adversarial networks, and stochastic differential equations to solve complex multi-agent market equilibrium models. However, its primary focus is on algorithmic formulation, approximation guarantees, and numerical experiments rather than empirical backtesting on real-world financial datasets.

Research Flowchart

  flowchart TD
    A[Research Goal: Solve Continuous-Time Market Equilibria with Frictions & Multiple Agents] --> B[Inputs: Asset Models, Trading Costs, Friction Parameters & Agent Utility Functions]
    B --> C[Methodology: Generative Deep RL with Adversarial GAN-style Training]
    C --> D[Core Process: Stabilize Dynamics via Reinforcement Link Feedback System]
    D --> E[Core Process: Decouple Complex Equilibrium System with Theoretical Guarantees]
    E --> F[Outcomes: Stable Convergence & Predictions on Endogenous Asset Returns/Volatilities]

Cited References (56)

  1. Asset pricing with liquidity risk (2005)
  2. Can a financial transaction tax prevent stock price booms? (2015)
  3. Optimal execution of portfolio transactions (2001)
  4. Optimal execution with nonlinear impact functions and trading-enhanced risk (2003)
  5. Direct estimation of equity market impact (2005)
  6. Option hedging with smooth market impact (2016)
  7. Neural Operators Can Play Dynamic Stackelberg Games — → 2411.09644
  8. Universal approximation bounds for superpositions of a sigmoidal function (1993)
and 48 more
  1. Asymptotics for small nonlinear price impact: A PDE approach to the multidimensional case (2021)
  2. Approximation rates for deep calibration of (rough) stochastic volatility models (2024)
  3. The Dynamic Properties of Financial-Market Equilibrium with Trading Fees (2019)
  4. Deep hedging (2019)
  5. A Dynamic Equilibrium Model of Liquidity Risk (2025)
  6. Deep Q-learning for Nash equilibria: Nash-DQN (2022)
  7. Portfolio selection with transactions costs (1976)
  8. Portfolio selection with transaction costs (1990)
  9. Optimal nonlinear approximation (1989)
  10. Liquidity regimes and optimal dynamic asset allocation (2020)
  11. Incomplete-Market Equilibria Solved Recursively on an Event Tree (2012)
  12. An exact solution to a dynamic portfolio choice problem under transactions costs (1991)
  13. Designing universal causal deep learning models: The case of infinite-dimensional dynamical systems from stochastic analysis (2022) — link
  14. Dynamic trading with predictable returns and transaction costs (2013)
  15. Dynamic portfolio choice with frictions (2016)
  16. Asset pricing with general transaction costs: Theory and numerics (2021)
  17. Approximation bounds for random neural networks and reservoir systems (2023)
  18. Deep ReLU neural networks overcome the curse of dimensionality for partial integrodifferential equations (2023)
  19. Generative adversarial networks (2014) — link
  20. Nonlinear price impact and portfolio choice (2020)
  21. Solving high-dimensional partial differential equations using deep learning (2018)
  22. Convergence of the deep BSDE method for coupled FBSDEs (2020)
  23. Evaluating the effects of incomplete markets on risk sharing and asset pricing (1996)
  24. Equilibrium asset pricing with transaction costs (2021)
  25. Multilayer feedforward networks are universal approximators (1989)
  26. The general structure of optimal investment and consumption with small transaction costs (2017)
  27. Universal approximation with deep narrow networks (2020)
  28. Universal approximation theorems for differentiable geometric deep learning (2022)
  29. Approximation Rates and VC -Dimension Bounds for (P)Re LU MLP Mixture of Experts (2025) — link
  30. Master curve for price-impact function (2003)
  31. Deep network approximation for smooth functions (2021)
  32. Closed-loop Nash competition for liquidity (2023)
  33. Deep vs. shallow networks: An approximation theory perspective (2016)
  34. Trading with small price impact (2017)
  35. An equilibrium model for the cross section of liquidity premia (2023)
  36. Dynamic portfolio choice with intertemporal hedging and transaction costs (2023)
  37. Universal approximation property of random neural networks (2023) — link
  38. Price impact equilibrium with transaction costs and TWAP trading (2022)
  39. Optimal approximation of piecewise smooth functions using deep ReLU neural networks (2018)
  40. Mathematical theory of deep learning (2024) — link
  41. Continuous-time stochastic control and optimization with financial applications (2009)
  42. Equilibrium asset pricing with transaction costs (2020)
  43. Deep learning algorithms for hedging with frictions (2023)
  44. Optimal investment and consumption with transaction costs (1994)
  45. Homogenization and asymptotics for small transaction costs (2013)
  46. Noncompact uniform universal approximation (2024)
  47. Error bounds for approximations with deep ReLU networks (2017)
  48. Deep network approximation: Beyond ReLU to diverse activation functions (2024)

1 of 56 cited works are in the corpus; resolved via arXiv source bundles, Semantic Scholar and OpenAlex. Corpus-wide: 2518 linked papers.