--- title: Portfolio & Allocation url: https://www.ml-quant.com/topics/portfolio-allocation/ site: ML-Quant (https://www.ml-quant.com) updated: 2026-10-09 license: Summaries CC BY 4.0; links go to the original sources index: https://www.ml-quant.com/llms.txt --- # Portfolio & Allocation Portfolio construction, allocation, rebalancing and risk budgeting, from Markowitz to deep RL. 628 papers featured; 51 in the last 12 months. Papers featured per quarter: 2023 Q2 34, 2023 Q3 67, 2023 Q4 83, 2024 Q1 66, 2024 Q2 76, 2024 Q3 60, 2024 Q4 42, 2025 Q1 70, 2025 Q2 58, 2025 Q3 17, 2025 Q4 26, 2026 Q1 1, 2026 Q2 2, 2026 Q3 7, 2026 Q4 19 ## Most cited - [Missing values handling for machine learning portfolios](https://www.ml-quant.com/papers/arxiv/2207.13071/): 35 citations. The study shows that using cross-sectional means for simple imputation is effective in dealing with missing values in machine learning-constructed portfolios, as complex imputations can cause underperformance due to estimation noise. - [Technical Note - An Unexpected Stochastic Dominance: Pareto Distributions, Dependence, and Diversification](https://www.ml-quant.com/papers/arxiv/2208.08471/): 27 citations. The research suggests that diversifying super-Pareto losses increases portfolio risk, discouraging risk sharing in market equilibrium. - [Dynamic asset allocation with asset-specific regime forecasts](https://www.ml-quant.com/papers/arxiv/2406.09578/): 24 citations. The article introduces a new framework that enhances multi-asset portfolio construction by creating custom regime forecasts for each asset, proven effective through a practical study on a multi-asset portfolio. - [MAD risk parity portfolios](https://www.ml-quant.com/papers/arxiv/2110.12282/): 23 citations. Features & Performance: A study using the Mean Absolute Deviation (MAD) to measure risk in the Risk Parity (RP) model found that RP strategies typically perform between minimum risk and equally weighted strategies. - [AlphaSAGE: Structure-Aware Alpha Mining via GFlowNets for Robust Exploration](https://www.ml-quant.com/papers/arxiv/2509.25055/): 21 citations. Alpha Mining: AlphaSAGE, a new framework for automated alpha mining in quantitative finance, uses a structure-aware encoder and Generative Flow Networks to overcome challenges and outperforms existing methods in creating a diverse and predictive portfolio of alphas. - [Developing A Multi-Agent and Self-Adaptive Framework with Deep Reinforcement Learning for Dynamic Portfolio Risk Management](https://www.ml-quant.com/papers/arxiv/2402.00515/): 18 citations. The piece introduces a multi-agent and self-adaptive framework (MASA) for portfolio management, using reinforcement learning to balance returns and risks, providing market trend feedback and outperforming other similar approaches. - [Risk Budgeting Allocation for Dynamic Risk Measures](https://www.ml-quant.com/papers/arxiv/2305.11319/): 17 citations. Risk budgeting allocation approach developed using dynamic risk contributions and deep learning. - [Explainable post hoc portfolio management financial policy of a Deep Reinforcement Learning agent](https://www.ml-quant.com/papers/arxiv/2407.14486/): 16 citations. A new Explainable Deep Reinforcement Learning (XDRL) method for portfolio management has been developed, combining Proximal Policy Optimization with explainable techniques for better transparency in investment predictions. - [Loss-Versus-Rebalancing under Deterministic and Generalized block-times](https://www.ml-quant.com/papers/arxiv/2505.05113/): 15 citations. A study reveals that constant block intervals in blockchain settings provide the best protection against arbitrage for Automated Market Makers' liquidity providers, using random walk theory. - [Sparse spanning portfolios and under-diversification with second-order stochastic dominance](https://www.ml-quant.com/papers/arxiv/2402.01951/): 15 citations. A new method for estimating sparse second-order stochastic spanning suggests no advantage in expanding a sparse opportunity set beyond 45 assets, with the best sparse portfolio investing in 10 sectors. - [Sparse spanning portfolios and under-diversification with second-order stochastic dominance](https://www.ml-quant.com/papers/ssrn/4713517/): 15 citations. The study explores whether relaxing sparsity constraints on portfolios enhances investment opportunities, finding no benefit from expanding a sparse opportunity set beyond 45 assets. - [On Optimal Tracking Portfolio in Incomplete Markets: The Reinforcement Learning Approach](https://www.ml-quant.com/papers/arxiv/2311.14318/): 15 citations. The paper uses capital injection to solve an optimal tracking portfolio problem in incomplete market models, showcasing the q-learning algorithm's performance. ## Latest - [Forecast Accuracy Is Not Trading Profit: Evolving Small Recurrent Networks for Stock Return Prediction](https://www.ml-quant.com/papers/arxiv/2610.07825/) (2026-10-09): Tiny neuroevolved recurrent networks rank first on both forecast accuracy and daily long-short returns across four portfolios, outperforming larger transformers and proving that forecast accuracy does not guarantee trading profit. - [Expected Utility Regret Rule: Minimax and Bayes Optimal Portfolio Choice](https://www.ml-quant.com/papers/arxiv/2610.02290/) (2026-10-09): Proposes a portfolio rule that simultaneously selects class and estimates weights to minimize expected utility regret, attaining both minimax and Bayes lower bounds without requiring prior information. - [MemTrial: Learning When to Trust Memory in LLM Portfolio Agents](https://www.ml-quant.com/papers/arxiv/2610.11732/) (2026-10-09): MemTrial uses factorial design to isolate what each memory contributes to portfolio decisions by crediting experiences with their marginal effect rather than shared market moves. - [Credal Machine Learning for Risk-Averse Decision Making](https://www.ml-quant.com/papers/arxiv/2610.12115/) (2026-10-09): A method represents epistemic uncertainty with credal sets and combines them with a decision rule that avoids catastrophic predictions under conditional value-at-risk minimization. - [Conditional value-at-risk under reward-penalty mechanism with applications to robust portfolio management](https://www.ml-quant.com/papers/arxiv/2610.09246/) (2026-10-09): Derives a closed-form expression for worst-case conditional value-at-risk under reward-penalty mechanisms and distribution uncertainty, enabling optimal portfolio allocations that balance losses with downside risk. - [Optimal Investment to Reach a Financial Goal: A Stochastic Control Framework](https://www.ml-quant.com/papers/arxiv/2610.10046/) (2026-10-09): A framework characterizes optimal investment policies for reaching financial goals by deadline, revealing that policies can decrease in asset drift and need not converge to risk-free allocation. - [Robust enhanced index tracking portfolio selection under distributional uncertainty](https://www.ml-quant.com/papers/arxiv/2610.04221/) (2026-10-09): Robust optimization models minimize worst-case loss and downside risk for enhanced indexing under distributional uncertainty, beating benchmarks and standard robust approaches in out-of-sample wealth. - [The Evolution of Quantitative Equity Investing: Half a Century of Alpha Research and Institutional Adaptation](https://www.ml-quant.com/papers/ssrn/7566701/) (2026-10-09): Surveys fifty years of quantitative investing as a history of changing implementation boundaries from diversification through machine learning, emphasizing alpha decay and the gap between research and production systems. - [Volatility Targeting, Variance Drain, and the Financialization Effect: A Century of Evidence](https://www.ml-quant.com/papers/ssrn/7585278/) (2026-10-09): Analyzing 100 years of daily data, the study finds volatility scaling harmed returns before 1985 but improved them after, a shift attributed to derivatives markets and deregulation, with walk-forward Sharpe ratio of 1.01. - [A Multi-Strategy Quantitative Portfolio](https://www.ml-quant.com/papers/ssrn/7561158/) (2026-10-09): Eleven systematic strategies across equities, futures, and currencies achieve a 1.91 Sharpe ratio net of costs from January 2012 to June 2026, with 23.6% compound annual return. - [Beliefs That Predict Returns and Beliefs That Attract Flows: Policy Insights and Sentiment Catering in Mutual Funds](https://www.ml-quant.com/papers/repec/nbr-nberwo-35528/) (2026-10-09): The study shows that mutual fund managers' policy beliefs predict market returns and explain positive alphas, while sentiment beliefs attract flows but lack predictive power. - [Optimal Currency Strategies Under Deviations From Interest Parity](https://www.ml-quant.com/papers/repec/nbr-nberwo-35498/) (2026-10-09): Extending Campbell et al., the analysis shows empirically measured deviations from interest parity reverse optimal currency demands for emerging-market investors. - [Admissible Portfolio Optimization: Information Constraints, Conditional Efficient Frontiers, and the Price of Causal Identification](https://www.ml-quant.com/papers/arxiv/2610.00147/) (2026-10-02): The research makes conditioning information a decision variable subject to constraints, solving how look-ahead bias and information availability affect portfolio choice and causal identification. - [Multiperiod bond portfolio optimization with transaction costs using a Markov Decision process](https://www.ml-quant.com/papers/arxiv/2609.38765/) (2026-10-02): The study develops a Markov decision process framework for dynamic bond portfolio management that balances yield and interest-rate risk subject to transaction costs. - [Taming the Greeks: Option Portfolios with Inductive Biases](https://www.ml-quant.com/papers/arxiv/2609.33767/) (2026-10-02): A deep learning framework for options trading enforces portfolio-level delta neutrality and other risk constraints during training, improving risk-adjusted returns with lower directional exposure. - [When Stress Tests Miss the Risk: Statistical Scenario Analysis for Financial Portfolios](https://www.ml-quant.com/papers/arxiv/2609.37273/) (2026-10-02): Kernel scenario analysis estimates quantile predictions for portfolio stress-test gains, with online calibration to capture dependence between stressed and unstressed risk factors. - [The Capacity of Equity Anomalies](https://www.ml-quant.com/papers/ssrn/7524338/) (2026-10-02): Computing capacity for 35 anomalies reveals that high-alpha long legs support less than $2 billion before trading costs eliminate excess returns, while lower-alpha strategies support billions more. - [Less Is More: Forecast Granularity, Estimation Error, and Portfolio Choice](https://www.ml-quant.com/papers/ssrn/7547959/) (2026-10-02): Investors using machine-learning forecasts can achieve Sharpe ratios of 1.2 by adjusting the number of portfolio groups based on the forecast's information coefficient, beating standard decile sorts. - [Locking in the rate or staying flexible? Mortgage refinancing around an interest rate shock](https://www.ml-quant.com/papers/repec/boe-boeewp-023579/) (2026-10-02): The study finds that UK borrowers shifted toward two-year fixed mortgages despite higher pricing after the 2022 rate shock, seeking flexibility and rate protection rather than minimizing immediate costs. - [Active Portfolio Management in Concentrated Equity Markets](https://www.ml-quant.com/papers/arxiv/2609.27113/) (2026-09-25): Formulates a stochastic control problem for actively allocating between equal-weighted and market portfolios based on a diversity-dispersion model, outperforming passive strategies during market bubbles. - [Decision-Focused Learning for Mean-Variance Portfolio Optimization via KKT-Based Reformulation](https://www.ml-quant.com/papers/arxiv/2609.21427/) (2026-09-25): Proposes a KKT-based decision-focused learning method that trains mean-variance portfolio models by directly minimizing downstream portfolio loss while preserving all constraints. - [The Critical Line Algorithm and the Constrained LASSO: One Curve, Two Literatures](https://www.ml-quant.com/papers/arxiv/2609.25704/) (2026-09-25): Shows that mean-variance portfolio selection and the constrained LASSO trace identical piecewise-linear solution paths, mapping their parametrizations exactly. - [Optimal Investment and Consumption in Financial Markets with Integrated Variance Clocks](https://www.ml-quant.com/papers/arxiv/2609.26349/) (2026-09-25): Characterizes optimal consumption and investment strategies in markets with stochastic volatility clocks using infinite-horizon backward SDEs, extending to rough and hyper-rough regimes. - [Welcome to the Factor Zoo: Where Mutual Fund Alpha Hides](https://www.ml-quant.com/papers/ssrn/7508299/) (2026-09-25): Using factor selection, the study finds mean active alpha of plus 9 basis points monthly for mutual funds, reversing the no-alpha conclusion when benchmarks are tailored to each fund. - [Data-Driven Minimax-Regret Portfolio Optimization under Tail-Risk Ambiguity](https://www.ml-quant.com/papers/ssrn/7486600/) (2026-09-25): The research proposes a data-driven portfolio method that blends tail-risk models and projects onto valid mixtures, providing bounds on Expected Shortfall regret without Wasserstein assumptions. - [HKC05 - Household Portfolios, Corporate Leverage, and the Supply Side of Monetary Policy](https://www.ml-quant.com/papers/repec/cxv-wpaper-2602/) (2026-09-25): Corporate leverage affects how monetary tightening transmits to the real economy: equity holders lose wealth while safe-asset holders are cushioned, raising the sacrifice ratio. - [Causal PDE-Control Models for Dynamic Portfolio Optimization with Latent Drivers](https://www.ml-quant.com/papers/arxiv/2509.09585/) (2026-04-16): Causal PDE-Control Models (CPCMs) offer a strong and clear framework for portfolio allocation that combines causal factors and complex filtering, outperforming standard econometric and machine-learning techniques. - [From Core to Periphery? Assessing Remote Works Potential to Rebalance EU Regional Development](https://www.ml-quant.com/papers/arxiv/2604.08252/) (2026-04-16): Remote work after the pandemic is causing people to move within cities for better quality of life, rather than relocating to rural areas. - [DeePM: Regime-Robust Deep Learning for Systematic Macro Portfolio Management](https://www.ml-quant.com/papers/arxiv/2601.05975/) (2026-01-16): Deep Learning for Portfolio Management: DeePM uses deep learning to improve macro portfolio management, delivering better risk-adjusted returns than traditional methods across various economic conditions. - [Vaccine Innovation Funding Strategy](https://www.ml-quant.com/papers/ssrn/4480682/) (2025-12-28): A portfolio approach to drug development may improve investment returns and speed up vaccine creation.