ABSTRACT Inverse design of functional materials—using target performance to guide optimal parameters—provides a powerful alternative to traditional forward methods, especially for complex, high‐dimensional problems. Advances in machine learning (ML) enhance its feasibility through fast surrogate modeling, efficient design‐space exploration, and direct mapping from desired properties to material solutions. This review presents a unified overview of ML‐driven inverse design methodologies, covering topology optimization, direct inverse mapping, and hybrid frameworks. We analyze key ML models, optimization algorithms, and adaptive schemes that tackle challenges including data scarcity and coupled physical constraints. Focusing on diverse functional materials, we highlight and illustrate how ML‐based inverse design is accelerating innovation across diverse classes of materials by rapid generation of microstructures and geometries tailored to specific functionalities, including mechanical and architected materials, acoustic and thermal metamaterials, optical materials, energy functional materials, biomedical and chemical materials. Finally, we outline key challenges and future directions toward autonomous, physics‐integrated, and generative pipelines for advanced functional materials. This review aims to provide a unified foundation for ML‐based inverse design and to guide the development of intelligent discovery pipelines for advanced materials.
Lin et al. (Thu,) studied this question.