Enhanced Power Estimation of Digital VLSI Circuits Using Machine Learning

Accurate power estimation remains a critical component of VLSI design, enabling valuable performance optimization. Traditional estimation techniques are not suitable for the early stages of design abstraction where speedy iteration is vital. In such a case, machine learning (ML) is a promising option with the ability to learn patterns and deliver quick, data-driven predictions. In this study, there is a comprehensive comparative evaluation of five advanced ML regressors Linear Regression, Decision Tree Regressor, Random Forest Regressor, Gradient Boosting Regressor, and Xtreme Gradient Boosting (XGBoost)-on dynamic power estimation in digital VLSI circuits. The dataset was produced through features drawn at synthesis level from designs which were processed by the Quartus Prime Design Suite to ensure usability and relevance. Model performance was analyzed through Root Mean Squared Error (RMSE) and R-squared (R<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup>) scores in order to gauge predictive accuracy and generalizability. Among all the models, XGBoost consistently outperformed, for which the RMSE was 2.04 and the R<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> was 0.9888, indicating high accuracy and reliability. Compared to results earlier reported in other research, the method in this work finds an impressive estimation accuracy enhancement of about 15.2% validating the potential of ML-based techniques in speeding up and enhancing power modeling efficiency and accuracy in contemporary VLSI designing processes.