arXiv Paper Analyzes SOMA and Differential Evolution Proposal Geometry

There’s a new study out on arXiv (2607.29228) that introduces an interesting way to analyze how the Self-Organizing Migrating Algorithm (SOMA) works compared to Differential Evolution (DE). This research focuses on a unique operator-selection factorization that separates variations not tied to objectives from boundary repairs and fitness-based selections. It reveals that the typical SOMA proposal behaves in an affine manner within its search space and is specifically linear when looking at an enhanced migrant-leader state. The authors also provide clear interpretations for various operations in relation to the leader and derive important formulas for proposal mean, covariance, and distance metrics while using Bernoulli perturbation masks. You can check it out at https://arxiv.org/abs/2607.29228.