Search papers, labs, and topics across Lattice.
This paper introduces ERRLESS, a novel approach to Bayesian symbolic regression that employs maximum-entropy reinforcement learning to sample from the posterior distribution of algebraic expressions. By utilizing a neural policy to sequentially construct expressions represented as abstract syntax trees, ERRLESS not only captures epistemic uncertainty but also generates interpretable and concise expressions. The method demonstrates competitive performance on the Feynman benchmark, achieving a high coefficient of determination compared to traditional sequential Monte Carlo baselines, underscoring the advantages of a Bayesian framework in this context.
ERRLESS reveals that incorporating Bayesian principles into symbolic regression can yield more interpretable and accurate models, outperforming traditional methods in uncertainty quantification.
Symbolic regression is the problem of finding an algebraic expression describing a stochastic dependence of a target variable on a set of inputs. Unlike forms of regression that fit parameters assuming a fixed model structure, symbolic regression is a search problem over the space of expressions, represented, for example, as abstract syntax trees using a library of operators. Symbolic regression is typically used in settings with limited, noisy data in the natural sciences. However, searching for a single best-fitting expression fails to capture the epistemic uncertainty about the expression, which motivates a Bayesian perspective that enables uncertainty quantification and specification of natural priors to constrain the search space. In this work, we propose ERRLESS (Entropy-Regularized Reinforcement Learning for Expression Structure Sampling), a scalable approach for sampling from the posterior distribution over expressions given data using maximum-entropy reinforcement learning. ERRLESS learns a neural policy that constructs expressions sequentially by building up their abstract syntax trees. At convergence, the policy samples expressions from the posterior. At test time, expressions can be sampled by rollouts of this policy. We demonstrate that ERRLESS achieves competitive results on the Feynman benchmark while producing short and interpretable expressions. Additionally, we demonstrate that the mean of the posterior predictive approximated by ERRLESS achieves a high coefficient of determination ($R^2$) compared to an SMC baseline, highlighting the benefits of the Bayesian perspective in symbolic regression.
Mila