Data-efficient Machine Learning (DEML), or surrogate modeling enables efficient learning, modeling, and optimization in engineering fields where high-quality data is limited or costly to obtain. DEML can streamline the creation of digital twins and predictive models based on small-scale datasets, which is especially beneficial for industries where data acquisition is challenging. By reducing the need for large datasets, DEML minimizes the time, resources, and engineering efforts needed to implement AI solutions in industrial settings.

A key application of DEML is in design optimization for complex, computationally intensive simulations, such as physics-based simulations used in fields like electromagnetics or computational fluid dynamics. For instance, DEML can optimize the design and electrical characteristics of components like microstrip filters, enhancing both their topology and functionality with fewer computational resources.