Entity · paper

Knowledge- and Gradient-Guided Reinforcement Learning for Parametrized Action Markov Decision Processes

paperactiveprovisionalknowledge-and-gradient-guided-reinforcement-learning-for-parametrized-action-markov-decision-processes-8447d20e·1 events·first seen 4d ago

Aliases: Knowledge- and Gradient-Guided Reinforcement Learning for Parametrized Action Markov Decision Processes

Co-occurring entities

Datalog KGRL

More like this (12)

Geometric Action Model for Robot Policy Learning Gradient-Guided Reward Optimization Improving LLM-Generated Process Model Quality Through Reinforcement Learning: The Role of Reward Function Design Entropy-Regularized Reinforcement Learning Using Reward Uncertainty to Induce Diverse Behaviour in Reinforcement Learning Policy Gradient Methods Evolved Policy Gradients General Preference Reinforcement Learning Multi-Agent Reinforcement Learning from Delayed Marketplace Feedback for Objective-Weight Adaptation in Three-Sided Dispatch Q-learning Mask-Aware Policy Gradients for Diffusion Language Models Reinforcement Learning with Metacognitive Feedback

Recent events (1)

4arXiv · cs.AI·4d ago·source ↗

KGRL: Neuro-symbolic reinforcement learning for parametrized action spaces using Datalog knowledge bases

A new arXiv preprint introduces KGRL (Knowledge- and Gradient-Guided Reinforcement Learning), a neuro-symbolic algorithm for Parametrized Action Markov Decision Processes (PAMDPs) where each decision involves both a symbolic action and continuous numerical parameters. KGRL uses a Datalog knowledge base to prune infeasible actions and constrain parameter spaces, then applies gradient-based refinement to estimate optimal parameters during training and deployment. The approach improves sample efficiency and episodic return over state-of-the-art PAMDP baselines, and provides local procedural explanations for its decisions.

AI Safety Research Agent and Tool Ecosystem Datalog KGRL Knowledge- and Gradient-Guided Reinforcement Learning for Parametrized Action Markov Decision Processes