Paper: arXiv 2402.16118

Authors: Bruno Gašperov, Marko Đurasević, Domagoj Jakobovic

Abstract

The majority of standard approaches to financial portfolio optimization (PO) are based on the mean-variance (MV) framework. Given a risk aversion coefficient, the MV procedure yields a single portfolio that represents the optimal trade-off between risk and return. However, the resulting optimal portfolio is known to be highly sensitive to the input parameters, i.e., the estimates of the return covariance matrix and the mean return vector. It has been shown that a more robust and flexible alternative lies in determining the entire region of near-optimal portfolios. In this paper, we present a novel approach for finding a diverse set of such portfolios based on quality-diversity (QD) optimization. More specifically, we employ the CVT-MAP-Elites algorithm, which is scalable to high-dimensional settings with potentially hundreds of behavioral descriptors and/or assets. The results highlight the promising features of QD as a novel tool in PO.

Complexity vs Empirical Score

  • Math Complexity: 7.5/10
  • Empirical Rigor: 6.0/10
  • Quadrant: Holy Grail — high math complexity, high empirical rigor

Why this score: The paper employs advanced mathematical concepts from optimization theory and evolutionary computation (quality-diversity, CVT-MAP-Elites), resulting in high math complexity. It also demonstrates empirical rigor through computational experiments, algorithmic implementation details, and discussion of real-world factors like transaction costs, placing it above the threshold for practical application.

Research Flowchart

  flowchart TD
  A["Research Goal<br>Find diverse near-optimal portfolios<br>beyond single-solution MV framework"] --> B{"Key Methodology"}
  
  B --> C["Data Inputs<br>Asset returns & covariance matrix"]
  C --> D["Algorithm: CVT-MAP-Elites<br>Quality-Diversity Optimization"]
  
  D --> E["Computational Process<br>Divide behavioral space into cells<br>Iterate through solutions<br>Archive best per cell"]
  
  E --> F["Key Findings<br>1. Scalable to high dimensions<br>2. Diverse set of near-optimal portfolios<br>3. Robust alternative to MV sensitivity"]
  
  style A fill:#e1f5fe
  style B fill:#fff3e0
  style C fill:#f3e5f5
  style D fill:#e8f5e8
  style E fill:#fce4ec
  style F fill:#e8f5e8

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 novel quality‑diversity optimization method (CVT‑MAP‑Elites) to discover a diverse set of near‑optimal portfolios.

Source sentence
we present a novel approach for finding a diverse set of such portfolios based on quality-diversity (QD) optimization.

Assumes 5

  • Short selling is not permitted in the portfolio formulation. model
    Source sentence
    For simplicity, short selling is not permitted, but it can be easily integrated into the framework if needed, by relaxing the wi ≥0 constraint.
  • Behavior space niches are created using centroidal Voronoi tessellation (CVT). model
    Source sentence
    In both cases, centroidal Voronoi tessellation (CVT) is used to partition the BS into niches.
  • A risk‑aversion coefficient is given to compute the mean‑variance efficient frontier. model
    Source sentence
    Given a risk aversion coefficient, the MV procedure yields a single portfolio that represents the optimal trade‑off between risk and return.
  • Investors select the final portfolio in a second, a‑posteriori phase from the near‑optimal set. execution
    Source sentence
    In the second phase (the a posteriori analysis), the investor selects the final portfolio from the sub‑space of near‑optimal portfolios.
Show 1 more
  • Transaction costs can be considered ad‑hoc but are not built into the optimization objective. execution
    Source sentence
    Simultaneously, it permits the ad‑hoc consideration of market frictions (e.g., transaction costs) and other regulatory, liquidity, and risk concerns.

Data

No dataset named in the text.

Costs

Partial Costs appear only as a robustness check, a flat fee without spread or impact, or outside the headline numbers.

Transaction costs are mentioned as market frictions that can be considered ad‑hoc but are not incorporated into the headline performance results.

Source sentence
Simultaneously, it permits the ad‑hoc consideration of market frictions (e.g., transaction costs) and other regulatory, liquidity, and risk concerns.

Finds 2

  • All QD methods show high coverage of the near‑optimal region, indicating the capability of QD to discover diverse elite portfolios. statistical
    Source sentence
    The main findings, pointing to high coverage and QD‑scores, underscore the capacity of QD to serve as a novel instrument in the field of PO.
  • Experimental results collectively clearly point to the promising capabilities of QD as a novel tool for portfolio optimization. theoretical
    Source sentence
    the experimental results collectively clearly point to the promising capabilities of QD as a novel tool in the arsenal of modern PO practitioners.

Checked by others

No other paper in the archive has been found to support or contradict these findings (compared with 17 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 (42)

  1. Robust optimization by constructing near-optimal portfolios (2017)
  2. Robust Mean-Variance Optimization (2016)
  3. The sharpe ratio (1998)
  4. Mean-variance analysis in portfolio choice and capital markets (2000)
  5. On the sensitivity of mean-variance-efficient portfolios to changes in asset means: some analytical and computational results (1991)
  6. Honey, I shrunk the sample covariance matrix (2003)
  7. Window effect with Markov-switching GARCH model in cryptocurrency market (2021)
  8. Asset allocation: combining investor views with market equilibrium (1990)
and 34 more
  1. A generalized approach to portfolio optimization: Improving performance by constraining portfolio norms (2009)
  2. A practical guide to robust portfolio optimization (2021)
  3. Efficient asset management: a practical guide to stock portfolio optimization and asset allocation (2008)
  4. AutoAlpha: An efficient hierarchical evolutionary algorithm for mining alpha factors in quantitative investment (2020) — link
  5. Generative Meta-Learning Robust Quality-Diversity Portfolio — → 2307.07811
  6. Global optimization by basin-hopping and the lowest energy structures of Lennard-Jones clusters containing up to 110 atoms (1997)
  7. Novelty search and the problem with objectives (2011)
  8. Using centroidal voronoi tessellations to scale up the multidimensional archive of phenotypic elites algorithm (2017)
  9. Illuminating search spaces by mapping elites (2015) — link
  10. Robo-advisors and wealth management (2017)
  11. Improving optimization (1993)
  12. Portfolio Optimization-The Mean-Variance and CVaR Approach (2010)
  13. The challenges of managing large FX reserves: the case of Israel (2019)
  14. Support vector machines: theory and applications (2005)
  15. Robust Portfolio Selection with Near Optimal Centering (2019)
  16. On benchmarking functions for genetic algorithms (2001)
  17. Quality-diversity optimization: a novel branch of stochastic optimization (2021)
  18. Quality-Diversity Meta-Evolution: customising behaviour spaces to a meta-objective (2021) — link
  19. Adaptive rolling window selection for minimum variance portfolio estimation based on reinforcement learning (2020)
  20. Monte Carlo Elites: Quality-diversity selection as a multi-armed bandit problem (2021)
  21. Improving genetic programming with novel exploration-exploitation control (2019)
  22. A quantitative approach to tactical asset allocation (2007)
  23. Discovering the elite hypervolume by leveraging interspecies correlation (2018)
  24. Analysis of Near-Optimal Portfolio Regions and Polytope Theory (2018)
  25. Portfolio optimization with an envelope-based multi-objective evolutionary algorithm (2009)
  26. On Data-Driven Log-Optimal Portfolio: A Sliding Window Approach (2022)
  27. Hybrid bi-objective portfolio optimization with pre-selection strategy (2017)
  28. An introduction to evolutionary computation in finance (2008)
  29. Risk and probability premiums for CARA utility functions (1993)
  30. The capital asset pricing model: Theory and evidence (2004)
  31. Benchmarking quality-diversity algorithms on neuroevolution for reinforcement learning (2022) — link
  32. Quality diversity: A new frontier for evolutionary computation (2016)
  33. Evolution of swarm robotics systems with novelty search (2013)
  34. Qhull: Quickhull algorithm for computing the convex hull (2013)

1 of 42 cited works are in the corpus; resolved via arXiv source bundles, Semantic Scholar and OpenAlex. Corpus-wide: 2596 linked papers.