Multi-Factor Portfolio

Entropy-Guided Multiplicative Updates: KL Projections for Multi-Factor Target Exposures ArXiv ID: 2510.24607 “View on arXiv” Authors: Yimeng Qiu Abstract We introduce Entropy-Guided Multiplicative Updates (EGMU), a convex optimization framework for constructing multi-factor target-exposure portfolios by minimizing Kullback-Leibler divergence from a benchmark under linear factor constraints. We establish feasibility and uniqueness of strictly positive solutions when the benchmark and targets satisfy convex-hull conditions. We derive the dual concave formulation with explicit gradient, Hessian, and sensitivity expressions, and provide two provably convergent solvers: a damped dual Newton method with global convergence and local quadratic rate, and a KL-projection scheme based on iterative proportional fitting and Bregman-Dykstra projections. We further generalize EGMU to handle elastic targets and robust target sets, and introduce a path-following ordinary differential equation for tracing solution trajectories. Stable and scalable implementations are provided using LogSumExp stabilization, covariance regularization, and half-space KL projections. Our focus is on theory and reproducible algorithms; empirical benchmarking is optional. ...

Dynamic Factor Allocation Leveraging Regime-Switching Signals ArXiv ID: 2410.14841 “View on arXiv” Authors: Unknown Abstract This article explores dynamic factor allocation by analyzing the cyclical performance of factors through regime analysis. The authors focus on a U.S. equity investment universe comprising seven long-only indices representing the market and six style factors: value, size, momentum, quality, low volatility, and growth. Their approach integrates factor-specific regime inferences of each factor index’s active performance relative to the market into the Black-Litterman model to construct a fully-invested, long-only multi-factor portfolio. First, the authors apply the sparse jump model (SJM) to identify bull and bear market regimes for individual factors, using a feature set based on risk and return measures from historical factor active returns, as well as variables reflecting the broader market environment. The regimes identified by the SJM exhibit enhanced stability and interpretability compared to traditional methods. A hypothetical single-factor long-short strategy is then used to assess these regime inferences and fine-tune hyperparameters, resulting in a positive Sharpe ratio of this strategy across all factors with low correlation among them. These regime inferences are then incorporated into the Black-Litterman framework to dynamically adjust allocations among the seven indices, with an equally weighted (EW) portfolio serving as the benchmark. Empirical results show that the constructed multi-factor portfolio significantly improves the information ratio (IR) relative to the market, raising it from just 0.05 for the EW benchmark to approximately 0.4. When measured relative to the EW benchmark itself, the dynamic allocation achieves an IR of around 0.4 to 0.5. The strategy also enhances absolute portfolio performance across key metrics such as the Sharpe ratio and maximum drawdown. ...