In the last decades, field-programmable gate arrays (FPGAs) have become increasingly important to the electronics industry, offering higher performance and lower power consumption as transistor technology continues to scale skylight quilt pattern down.Machine learning (ML) algorithms have become pivotal in the electronic design automation (EDA) of FPGAs, enabling the learning of relationships between input and the desired output based on representative data properties rather than physical laws.As FPGA capacity expands, the EDA tools must also scale to handle larger, denser digital systems, and ML offers to fill the gap with resulting computational efficiency and improved solution quality.This yeti moth study reviews ML methods utilized in FPGA EDA, from the perspective of formulated problems, input space representation, learned mapping, and methods used to achieve that.
The work also presents the main ML methodologies and future challenges, serving as a roadmap for FPGA practitioners to navigate in the area of ML for FPGA EDA.