Location
MC 6460
Candidate
Peiyi Zheng | Applied Mathematics, University of Waterloo
Title
Transformer-Guided Symbolic Regression for Scientific Equation Recovery
Abstract
Symbolic regression aims to recover closed-form equations from observations, providing interpretable models for scientific discovery. Existing approaches struggle to combine flexible structural search with efficient inference. Search based methods can refine expression structure, but often rely on costly combinatorial optimization with random initialization. Pretrained neural models generate formulas almost instantly, but their predictions often contain symbolic errors. We introduce MOSAIC-SR, which uses a pretrained Transformer to propose multiple initial sketches. These sketches initialize searches in several promising regions, avoiding random starts in the vast expression space. Each search jointly recovers structure and constants through scale-aware constant optimization and local symbolic repair. Periodic migration transfers useful discoveries across searches. We evaluate MOSAIC-SR on several symbolic regression benchmarks, including datasets with extra dummy variables. Results show that MOSAIC-SR achieves competitive symbolic recovery and predictive accuracy across equations of varying difficulty. These findings demonstrate that learned priors can efficiently identify promising regions of expression space, while numerical optimization and symbolic search are essential for recovering exact equations.