Paper: arXiv 2511.21221

Authors: Kexin Wang, Xiaomeng Zhang, Xinyu Zhang

Abstract

Recognizing that asset markets generally exhibit shared informational characteristics, we develop a portfolio strategy based on transfer learning that leverages cross-market information to enhance the investment performance in the market of interest by forward validation. Our strategy asymptotically identifies and utilizes the informative datasets, selectively incorporating valid information while discarding the misleading information. This enables our strategy to achieve the maximum Sharpe ratio asymptotically. The promising performance is demonstrated by numerical studies and case studies of two portfolios: one consisting of stocks dual-listed in A-shares and H-shares, and another comprising equities from various industries of the United States.

Complexity vs Empirical Score

  • Math Complexity: 7.5/10
  • Empirical Rigor: 6.0/10 (capped from 7.0: transaction costs not modelled; see penalty flags)
  • Quadrant: Holy Grail — high math complexity, high empirical rigor (effective empirical rigor capped from 7.0)

Why this score: The paper presents advanced theoretical mathematics including asymptotic Sharpe ratio optimization, model averaging, and detailed proofs, yet it is grounded in empirical application with real-world case studies (A-shares/H-shares, US industries) and simulation-based validation.

Research Flowchart

  flowchart TD
  A["Research Goal: Enhance Portfolio Optimization<br>using Cross-Market Information"] --> B["Data Collection<br>A-shares/H-shares & US Equities"]
  B --> C["Methodology: Transfer Learning Framework"]
  C --> D["Computational Process: Asymptotic Identification<br>& Selection of Informative Datasets"]
  D --> E["Process: Discard Misleading Info<br>Retain Valid Cross-Market Signals"]
  E --> F["Outcome: Portfolio Strategy with<br>Maximum Asymptotic Sharpe Ratio"]
  F --> G["Validation: Numerical & Case Studies<br>Confirm Promising Performance"]

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 transfer learning based portfolio strategy that leverages cross‑market information to improve target market performance.

Source sentence
we develop a portfolio strategy based on transfer learning that leverages cross‑market information to enhance the investment performance in the market of interest by forward validation.

Assumes 2

  • There exist constants and parameter sequences such that d^{ζ}h^{-1/2}=o(1) and certain convergence rates hold for estimated means and covariances. statistical
    Source sentence
    ... (i) dζh−1 2 = o(1); (ii) h 1 2d−ζ ∥bµm,τi −µ∗ m,τi ∥2= Op(1) and h 1 2d−ζ ∥bΣm,τi −Σ∗ m,τi ∥F= Op(1); (iii) h 1 2d−ζ ∥µ∗ m,τi+1 −µ∗ m,τi ∥2= O(1) and h 1 2d−ζ ∥Σ∗ m,τi+1 −Σ∗ m,τi ∥F= O(1).
  • The target mean and covariance estimators converge at rate h^{1/2}d^{-ζ}. statistical
    Source sentence
    Assumption 3. (i) h 1 2d−ζ ∥µ0,τi −µ∗ 0,τi ∥2= O(1); (ii) h 1 2d−ζ ∥Σ0,τi −Σ∗ 0,τi ∥F= O(1).

Data

  • dual‑listed A‑shares and H‑shares portfolio market-prices unknown
    stocks dual‑listed in A‑shares and H‑shares
    Source sentence
    case studies of two portfolios: one consisting of stocks dual‑listed in A‑shares and H‑shares, and another comprising equities from various industries of the United States.
  • U.S. equities portfolio market-prices unknown
    equities from various industries of the United States
    Source sentence
    ... another comprising equities from various industries of the United States.

Recognised sources: CRSP Compustat

Costs

Not modeled A trading or execution result is reported without costs.

Finds 3

  • The proposed TL strategy automatically excludes uninformative source datasets and asymptotically concentrates weights on informative ones. theoretical
    Source sentence
    Our strategy automatically excludes uninformative and misleading source datasets, asymptotically concentrating portfolio weights on the informative datasets...
  • The TL strategy asymptotically achieves the maximum attainable Sharpe ratio, with lower variance than the non‑transfer baseline when informative sources exist. theoretical
    Source sentence
    ... the variance of the Sharpe ratio under our strategy is asymptotically lower than that of the Non−transfer baseline strategy
  • Numerical studies and case studies demonstrate promising performance of the TL strategy. performance
    Source sentence
    The promising performance is demonstrated by numerical studies and case studies of two portfolios...

Checked by others

No other paper in the archive has been found to support or contradict these findings (compared with 22 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 (55)

  1. PORTFOLIO SELECTION (1952)
  2. Improved estimation of the covariance matrix of stock returns with an application to portfolio selection (2003)
  3. A well-conditioned estimator for large-dimensional covariance matrices (2004)
  4. Nonlinear shrinkage of the covariance matrix for portfolio selection: Markowitz meets goldilocks (2018)
  5. Covariance regularization by thresholding (2008)
  6. High dimensional covariance matrix estimation using a factor model (2008)
  7. Large covariance estimation through elliptical factor models (2015)
  8. Asset allocation: combining investor views with market equilibrium (1991)
and 47 more
  1. Portfolio selection with parameter and model uncertainty: A multi-prior approach (2007)
  2. Risk Reduction in Large Portfolios: Why Imposing the Wrong Constraints Helps (2003)
  3. A generalized approach to portfolio optimization: Improving performance by constraining portfolio norms (2009)
  4. Vast portfolio selection with gross-exposure constraints (2012)
  5. Vast volatility matrix estimation using high-frequency data for portfolio selection (2010)
  6. Optimal portfolio choice with parameter uncertainty (2007)
  7. Markowitz meets Talmud: A combination of sophisticated and naive diversification strategies (2011)
  8. Why Naive Diversification Is Not So Naive, and How to Beat It? (2024)
  9. The Sharpe ratio (1994)
  10. Optimal algorithms and intuitive explanations for Markowitz’s portfolio selection model and Sharpe’s ratio with no short-selling
  11. Sparse minimax portfolio and Sharpe ratio models (2022)
  12. Optimal portfolio choice using the maximum Sharpe ratio (2010)
  13. The large-sample distribution of the maximum Sharpe ratio with and without short sales (2016)
  14. Transfer learning for pedestrian detection (2013)
  15. LST: Ladder Side-Tuning for Parameter and Memory Efficient Transfer Learning
  16. Jointly modeling transfer learning of industrial chain information and deep learning for stock prediction (2022)
  17. Transfer Learning for Portfolio Optimization — → 2307.13546
  18. Multiple Regression Model Averaging and the Focused Information Criterion With an Application to Portfolio Choice (2019)
  19. Econometrics (2022)
  20. A Survey on Transfer Learning (2010)
  21. A Comprehensive Survey on Transfer Learning (2021)
  22. Improving financial trading decisions using deep Q-learning: Predicting the number of shares, action strategies, and transfer learning (2019)
  23. Cross validation based transfer learning for cross-sectional non-linear shrinkage: A data-driven approach in portfolio optimization (2024)
  24. Optimal Parameter-Transfer Learning by Semiparametric Model Averaging (2023)
  25. QuantNet: transferring learning across trading strategies (2022)
  26. Prediction Using Many Samples with Models Possibly Containing Partially Shared Parameters (2024)
  27. Least Squares Model Averaging (2007)
  28. Model averaging in economics: An overview (2015)
  29. Improved estimation of the covariance matrix of stock returns with an application to portfolio selection (2003)
  30. Nonlinear Shrinkage of the Covariance Matrix for Portfolio Selection: Markowitz Meets Goldilocks (2017)
  31. Approaching Mean-Variance Efficiency for Large Portfolios (2018)
  32. Option Return Predictability with Machine Learning and Big Data (2023)
  33. Earnings Extrapolation and Predictable Stock Market Returns (2025)
  34. Mutual Fund Performance (1966)
  35. Common risk factors in the returns on stocks and bonds (1993)
  36. The distribution of realized stock return volatility (2001)
  37. Asset pricing: Revised edition (2009)
  38. How global is predictability? \text T he power of financial transfer learning (2024)
  39. Principal portfolios (2023)
  40. Risk of transfer learning and its applications in finance — → 2311.03283
  41. A berry-esséen bound for multivariate l-estimates with explicit dependence on dimension (2006)
  42. Dependence of the Berry-Esseen estimate on the dimension (1986)
  43. On non-uniform estimate of convergence rate in multidimensional central limit theorem with stable limit law (1989)
  44. On the rate of convergence in the multivariate CLT (1991)
  45. An estimate of the remainder term in the multidimensional central limit theorem (2006)
  46. On the dependence of the Berry-Esseen bound on dimension (2003)
  47. Central limit theorems for high dimensional dependent data (2024)

2 of 55 cited works are in the corpus; resolved via arXiv source bundles, Semantic Scholar and OpenAlex. Corpus-wide: 2596 linked papers.