MIRO: A Miró-Inspired Metaheuristic Optimization Algorithm

Abstract

This paper proposes MIRO (Miró-Inspired Rhythmic Orbits), a new population-based metaheuristic inspired by Joan Miró’s visual language: (i) constellation-like point fields, (ii) thin black linking lines, (iii) biomorphic deformations, and (iv) balance/equilibrium through repeated refinement. These artistic principles are mapped into three cooperating search operators—Splatter, Line-Weaving, and Equilibrium Balance—that jointly control exploration and exploitation. MIRO is evaluated on the convex test problem z = (x−3)^2 + (y−2)^2 and five classical benchmark functions (Sphere, Rosenbrock, Rastrigin, Ackley, Griewank). A pure NumPy + Matplotlib implementation is provided (no scikit-learn, no TensorFlow), producing colorful graphical outputs and reporting optimal points.