Abstract
Markowitz’s criterion aims to balance expected return and risk when optimizing the portfolio. The expected return level is usually fixed according to the risk appetite of an investor, then the risk is minimized at this fixed return level. However, the investor may not know which return level is suitable for her/him and the current financial circumstance. It motivates us to find a novel approach that adaptively optimizes this return level and the portfolio at the same time. It not only relieves the trouble of deciding the return level during an investment but also gets more adaptive to the ever-changing financial market than a subjective return level. In order to solve the new model, we propose an exact, convergent, and efficient Krasnoselskii-Mann Proximity Algorithm based on the proximity operator and Krasnoselskii-Mann momentum technique. Extensive experiments show that the proposed method achieves significant improvements over state-of-the-art methods in portfolio optimization. This finding may contribute a new perspective on the relationship between return and risk in portfolio optimization.
Introduction
Modern Portfolio Theory, pioneered by Harry Markowitz, has long relied on the Mean-Variance (MV) framework. In this paradigm, investors traditionally select a target return level ($\rho$) and then minimize the portfolio's variance to meet that target. However, this approach introduces a significant "subjectivity gap." In practice, investors often struggle to determine a realistic or optimal return level amidst volatile market conditions. Setting $\rho$ too high invites excessive risk, while setting it too low results in missed opportunities.
To address this, this research introduces the Markowitz Portfolio with Adaptive Expected Return Level (MPAERL) and a corresponding solver, the Krasnoselskii-Mann Proximity Algorithm (KMPA). By treating the return level as a variable to be optimized rather than a fixed constraint, the model dynamically adjusts to market shifts, relieving the investor of the burden of subjective target-setting.
The MPAERL Model: Beyond Fixed Constraints
The core innovation of the MPAERL model is the transition from a rigid return constraint to an adaptive optimization objective. Traditional models, such as the Sparse Standard Markowitz Portfolio (SSMP), treat the return level as a constant. MPAERL, however, optimizes the return level and portfolio weights simultaneously.
Furthermore, the model incorporates $\ell_1$-norm regularization (Lasso). This encourages "sparsity" in the portfolio, ensuring that capital is concentrated in a small subset of high-impact assets. This sparsity is crucial for practical trading, as it reduces transaction costs and simplifies portfolio management compared to "dense" portfolios that hold negligible positions in hundreds of assets.
The KMPA Algorithm: Efficiency through Momentum
Solving the MPAERL model requires a robust numerical approach. The authors propose the Krasnoselskii-Mann Proximity Algorithm (KMPA), which leverages two key mathematical concepts:
- Proximity Operators: These are essential for handling non-smooth functions, such as the $\ell_1$-norm penalty used for sparsity.
- Krasnoselskii-Mann (KM) Momentum: This fixed-point iteration technique accelerates convergence.
By combining these elements, KMPA provides an "exact" and computationally efficient solver. Unlike heuristic methods, KMPA is mathematically guaranteed to converge to a stable solution, even when navigating complex inequality constraints and non-smooth penalties. This momentum-based framework ensures the algorithm remains stable and fast, making it suitable for high-dimensional financial data.
Key Insights and Results
The research demonstrates that breaking the fixed-return constraint allows the portfolio to "breathe" with the market. Instead of forcing the portfolio to hit a potentially arbitrary target on the efficient frontier, the MPAERL model helps navigate the frontier dynamically as market conditions evolve.
Experimental Findings: * Performance Gains: Extensive benchmarking against state-of-the-art (SOTA) methods shows significant improvements in risk-adjusted returns. * Reduced Investor Burden: By automating the selection of the return level, the model mitigates human bias and the difficulty of deciding on targets in uncertain environments. * Market Adaptability: The adaptive nature of the model allows it to respond more effectively to changing financial climates than static, subjective models.
Future Directions and Limitations
While KMPA represents a significant step forward, challenges remain regarding hyperparameter sensitivity, specifically for the regularization parameter ($\tau$) and momentum steps. Future work may explore:
- Alternative Risk Measures: Extending the adaptive framework to Conditional Value at Risk (CVaR) or other non-variance risk metrics.
- Broader Applications: The authors note that KMPA is a versatile solver for a broader class of two-term convex optimization models, suggesting potential applications in signal processing and image reconstruction.
Conclusion
The Krasnoselskii-Mann Proximity Algorithm offers a new perspective on the relationship between return and risk. By moving toward adaptive return levels, investors can achieve portfolios that are not only mathematically optimized but also resilient to the inherent unpredictability of global markets.
Summary updated