Research
My research focuses on developing novel approaches to combinatorial optimization and scientific machine learning problems, combining classical methods with modern AI techniques.
Tackling the No-Three-In-Line Problem
Comparing Generative AI and Classical Optimization Methods for the No-Three-In-Line Problem
Advanced research combining classical optimization methods (recursive search, ILP with Gurobi), deep learning (PatternBoost transformer), and reinforcement learning (PPO agent) to solve the No-Three-in-Line problem. The innovative UDE extension treats N3L as a physics problem with continuous dynamics, energy functionals penalizing collinearity, neural correction terms, and universal differential equation formulations.
HP Protein Folding (CP-SAT + PPO + Hybrid)
Hybrid Solver-Guided Transformer for Protein Structure Prediction
Comprehensive study of the HP protein folding model using multiple approaches: CP-SAT baseline optimization, DQN reinforcement learning agent.