Limited Perfect Monotonical Surrogates constructed using low-cost recursive linkage discovery with guaranteed output

This project is a very early-stage academic reference implementation (0 stars, 4 days old) associated with a research paper on evolutionary computation. Its 'defensibility' is currently near zero as it lacks a community, documentation for production use, or a proven track record outside of specific benchmark functions. The core innovation lies in moving from 'approximate' surrogates (like Kriging or Neural Networks) to 'perfect' surrogates that map the structural linkage of the objective function. While intellectually significant for researchers in Estimation of Distribution Algorithms (EDAs) or Genetic Algorithms, it currently serves as a niche tool for global optimization of expensive black-box functions. Frontier labs (OpenAI/Google) are unlikely to compete here directly, as this is deep operations research/evolutionary computation rather than general-purpose AI. The primary threat comes from established optimization frameworks like Optuna or BoTorch potentially incorporating similar decomposition techniques if they prove to be broadly superior to Gaussian Processes or Tree-structured Parzen Estimators.