Inverse problems and inverse design, aim to minimize different objectives under given constraints, for which gradient-based optimization is most commonly employed. Similarly, machine learning heavily relies on gradient-based optimization of other objective functions. The key distinction lies in machine learning aiming at generalizing beyond the studied, i.e., optimized problems. Nevertheless, attempts to exploit connections between the two methodologies have been initiated for a broad range of inverse problems in structural mechanics, ranging from Meta-/architected materials to structural optimization. This mini-symposium aims at further exploring the possibility of integrating, adapting, and further developing techniques of machine learning into challenges encountered in structural mechanics. Most important in these treatments are transparency and honesty in the methods' performances, ensuring realistic impressions of advances and potentials. Exemplary subtopics include:
• exploiting recent advances in machine learning for inverse problems or inverse design
• explainable or robust machine learning methods for optimization or surrogate modeling
• generative artificial intelligence for inverse design and metamaterials
• material parametrizations with neural networks
• surrogate models for substeps of the optimization chain
• complex physical phenomena, where current optimization frameworks are prohibitively time-consuming