Conceptual
Browse Instruct Curate

Noise-Resilient Symbolic Regression with Reinforcement Learning and Dynamic Gating

A reinforcement-learning approach to symbolic regression that recovers accurate closed-form expressions from high-noise data: a policy network emits expression tokens under a fitness reward, a noise-resilient gating module dynamically filters uninformative noisy inputs, and a mixed-path-entropy bonus enlarges exploration of the expression space, matching state-of-the-art accuracy on clean data while improving robustness on noisy data.