Author ORCID Identifier

https://orcid.org/0009-0003-2868-4977

Defense Date

2026

Document Type

Dissertation

Degree Name

Doctor of Philosophy

Department

Systems Modeling and Analysis

First Advisor

Dr. J. Paul Brooks

Second Advisor

Dr. Craig E. Larson

Abstract

Scientific discovery increasingly relies on identifying interpretable mathematical relation- ships within high-dimensional, noisy datasets. Within computational scientific discovery, Graffiti-like automated mathematical conjecturing systems generate symbolic candidate expressions that capture empirical relationships, but often produce an unmanageable number of candidates, shifting the computational bottleneck from generation to selection. This dissertation presents a dual approach to principled expression selection. First, we introduce the Structural Average Model (SAM), an optimization-based framework that identifies a “best” set of symbolic expressions by balancing robustness, fidelity, and parsi- mony. Second, we formulate the selection process as a Markov Decision Process (MDP) and employ reinforcement learning to optimize set-level interpretability, moving beyond simple additive objectives. We provide theoretical guarantees for the well-posedness and convergence of these methods. Additionally, we demonstrate the utility of these frameworks in forensic microbiome analysis, where the discovery of interpretable symbolic bounds for body fluid and mixture identification is the key to establishing courtroom admissibility and expert testimony. Finally, we demonstrate its application to graph theory, pushing the boundaries of current knowledge regarding the NP-hard independence number problem. This work advances computational scientific discovery by providing principled optimization and learning methodologies for selecting interpretable symbolic conjectures while preserving the central role of human investigation.

Rights

© The Author

Is Part Of

VCU University Archives

Is Part Of

VCU Theses and Dissertations

Date of Submission

8-5-2026

Available for download on Monday, August 04, 2031

Share

COinS