The Meritorious Effects of Machine Learning Algorithms in Assessment of Power Consumption by Arithmetic Circuits in IC Design

The Machine Learning models will either learn from the experimental measurements or the observed data through simulations that depict the physical problem in order to identify the correlation among the input variables and the outputs of interest that are generated by the system that is being investigated. These days, machine learning models are getting a lot of attention in a lot of different industries because of their capacity to model complicated systems by using historical data. In applications that call for real-time estimation of the variables, the utilisation of data-driven models has been shown to be effective and has been successfully proven. The study that is being presented makes use of six distinct machine learning algorithms, each of which possesses the capability to estimate the amount of power that is consumed by digital circuits, even in the absence of knowledge of the actual structure or interconnections of the circuits. In order to represent the algorithm, the number of gates that are involved in the design is selected as the input. The simulation is carried out with the assistance of the Orange Data Mining tool using the ISCAs'89 data collection. According to the findings, the AdaBoost algorithm performs better than any other method with an RMSE value of 0.503, when it comes to predicting the amount of power that is used by a digital circuit.