Code and Data for Manuscript: "Machine learning prediction of the convergence criterion for a topological invariant of finite non-Hermitian chains"

This archive contains the data tables, trained random-forest (RF) models, and Python scripts needed to reproduce the figures and numerical scores in the paper. The work concerns point-gap topology of non-Hermitian (Hatano-Nelson-type) lattice models, where a real-space topological invariant is evaluated on cropped finite chains, and random forests are used to predict the minimum crop length ell_star needed to recover the correct winding number.