Paper Overview
Field: Machine Learning Authors: Zeynep Yıldız, Satchit Chatterji, Ercan Atam, Erman Acar Published: 2026-11-12 arXiv: 2611.17329
Abstract
Institutional portfolios operate under investment policy statements that specify asset allocation constraints, often conditioned on market regime information. Deep reinforcement learning has been shown to optimize financial performance but typically does not enforce rule compliance, making it unsuitable for real-world deployment. Existing neurosymbolic methods such as Probabilistic Logic Policy Gradient (PLPG) offer constraint-satisfying reinforcement learning (RL) via differentiable symbolic supervision but assume a finite, discrete action space.
The authors present a PLPG-inspired continuous allocation method that estimates the probability that the policy's own executed action satisfies the investment mandate, using reparameterized Monte Carlo sampling, and adds it to the training objective. No post-hoc blocking mechanism is needed at inference time: compliance is internalized by the agent during training.
Key Results
Audited directly on the deterministic mean action, without any correction:
- Unconstrained RL: 96-98% violation rate on a six-rule, regime-switching investment mandate
- Reward-penalty baseline: 33.4% violation rate
- Proposed method: 4.8% violation rate
Takeaway
Continuous-action neurosymbolic policies can substantially internalize symbolic constraints, a meaningful step toward deploying RL-based portfolio management under real-world regulatory and mandate requirements.
--- *Auto-collected on 2026-11-14*