The inverse design of acoustic metamaterials where structures are engineered to meet specific acoustic performance targets has gained significant momentum through data-driven and generative approaches. Rather than relying solely on iterative simulations or manual tuning, modern algorithms enable design objectives to directly inform structure generation. Simulation datasets are commonly used to train models that efficiently map acoustic responses to physical geometries. Generative techniques, such as autoencoders, diffusion models, and frameworks inspired by large language models, have been applied to synthesize novel structures and broaden the design space. These models enable exploration beyond known datasets while preserving physical plausibility. Recent advances also incorporate interpretable machine learning to provide insight and feedback into the design process. When combined with physics-based constraints, these methods accelerate discovery and reduce computational overhead. However, several challenges remain, including generalization across frequency regimes, efficient high-fidelity data generation, and experimental validation. This talk will present recent progress in inverse design strategies for acoustic metamaterials, outline current challenges, and offer perspectives on future research directions in the field.
Krupali Donda (Wed,) studied this question.