Paper: arXiv 2602.16862

Authors: Andy Au

Abstract

We study entropy-regularized mean-variance portfolio optimization under Bayesian drift uncertainty. Gaussian policies remain optimal under partial information, the value function is quadratic in wealth, and belief-dependent coefficients admit closed-form solutions. The mean control is identical to deterministic Bayesian Markowitz feedback; entropy regularization affects only the policy variance. Additionally, this variance does not affect information gain, and instead provides belief-dependent robustness. Notably, optimal policy variance increases with posterior conviction $|m_t|$, forcing greater action randomization when mean position is most aggressive.

Complexity vs Empirical Score

  • Math Complexity: 8.5/10
  • Empirical Rigor: 1.5/10
  • Quadrant: Lab Rats — theoretically deep, empirically untested

Why this score: The paper presents a rigorous theoretical bridge between entropy-regularized control and Bayesian drift learning in continuous-time portfolio optimization, deriving clean closed-form solutions via PDE analysis. However, it is entirely theoretical with zero empirical validation or backtesting on market data. Its strength lies in its mathematical clarity and counterintuitive analytical findings regarding belief-dependent policy variance.

Research Flowchart

  flowchart TD
    A["Research Goal: Entropy-regularized mean-variance<br/>portfolio choice under Bayesian drift uncertainty"] --> B["Inputs: Wealth state, asset returns,<br/>and prior belief on drift parameter"]
    B --> C["Bayesian Filtering: Update belief<br/>state (posterior mean & variance)"]
    C --> D["Stochastic Control: Formulate HJB equation<br/>with entropy penalty"]
    D --> E["Analytical Derivation: Solve for value function<br/>& optimal Gaussian policy"]
    E --> F["Outcome 1: Separation Principle<br/>Mean control matches classical Markowitz;<br/>Entropy affects only policy variance"]
    E --> G["Outcome 2: Belief-Dependent Robustness<br/>Policy variance grows with conviction |m_t|;<br/>Does not alter information gain"]

Cited References (7)

  1. Portfolio Selection (1952) — link
  2. On Estimating the Expected Return on the Market: An Exploratory Investigation (1980) — link
  3. Continuous-Time Mean-Variance Portfolio Selection: A Stochastic LQ Framework (2000) — link
  4. Bayesian Learning for the M arkowitz Portfolio Selection Problem (2018) — link
  5. Continuous-Time Mean–Variance Portfolio Selection: A Reinforcement Learning Framework (2019) — link
  6. Continuous-Time Stochastic Control and Optimization with Financial Applications (2009) — link
  7. Reinforcement Learning in Continuous Time and Space: A Stochastic Control Approach (2019) — link

0 of 7 cited works are in the corpus; resolved via arXiv source bundles, Semantic Scholar and OpenAlex. Corpus-wide: 2518 linked papers.