Exploring Machine Learning for Electrical Behavior Prediction: The CMOS Inverter Case Study

With the advancement of integrated circuit man-ufacturing technology, a growing number of aspects must be considered during the electrical characterization of circuits in order to solve challenges such as the effect of process variability. This increases the characterization time due to the use of techniques based on exhaustive electrical simulations. Machine learning techniques are consistently being employed to assist digital design at many levels of abstraction with various successful applications. Thus, the main objective of this work is to evaluate machine learning regression algorithms as an alternative to ex-haustive electrical simulation in the cell characterization project. In this step, multiple linear regression, support vector regression, decision trees, and random forest algorithms are considered. This work presents the results of a first case study: an Inverter using bulk CMOS technology. Specifically, the energy values and propagation times of this circuit will be separately predicted. A comparative analysis is done for each dependent variable between the models in order to understand which is the best regression model for the task. The algorithm with the lowest cost function proved to be Random Forests, with a R2 above 98% for all predicted variables.