Research
Embedded Files
Samuel Paul Gyamfi's research focuses on the application of mathematical modeling and machine learning to systems analysis, optimization, and data-driven decision-making. His work bridges the gap between theoretical frameworks and real-world challenges in finance, insurance, and structural analysis.
Samuel’s interests extend to advanced probabilistic modeling, particularly in leveraging machine learning algorithms to enhance predictive accuracy in risk assessment and decision-making. With a background in applied mathematics and actuarial science, he explores novel methodologies to improve efficiency and reliability in areas such as claims processing, underwriting, and portfolio optimization.
He is particularly drawn to the intersection of optimization and machine learning, seeking innovative ways to develop adaptable models that can address diverse challenges in dynamic systems. His research also includes a focus on data analysis and statistical modeling, key tools for understanding complex patterns and behaviors in systems.
Published Papers
Antimicrobial Resistance and Tuberculosis Prevalence in Africa: A Public Health Concern (Download Paper)
News
Title: Heuristics to Learning – A Reinforcement Approach to Mathematical Conjecturing
Title: Heuristics to Learning – A Reinforcement Approach to Mathematical Conjecturing
On April 29, 2025, Samuel Paul Gyamfi delivered a graduate seminar at Virginia Commonwealth University (VCU), presenting his research on the application of reinforcement learning in mathematical conjecturing. His talk, titled “Heuristics to Learning: A Reinforcement Approach to Mathematical Conjecturing,” introduced a novel method for automating symbolic reasoning tasks by integrating reinforcement learning (RL) into existing mathematical frameworks.
Drawing on the limitations of the deterministic Dalmatian heuristic used in the CONJECTURING tool by Brooks et al., Samuel proposed a dynamic alternative that leverages RL to filter and evaluate symbolic expressions. By framing the selection of mathematical expressions as a sequential decision-making problem, his approach enables the learning agent to generalize policies that balance complexity, accuracy, and interpretability—even under noisy or incomplete data conditions.
The seminar was well received by faculty and fellow graduate students and offered a compelling vision for the future of automated reasoning in mathematics.
Stay Updated!
Samuel plans to keep this page updated with the evolution of his research and academic journey. Check back for the latest insights into his work and contributions!
Photos from the 2024 INFORMS Annual Meeting
Page updated
Google Sites
Report abuse