Random Forest Regression Model AugmenteD With KL Algorithm (RFRKL) To Minimize Power In FSM Synthesis

The logic synthesis domain has recently benefited from improvements in intelligent learning, particularly for low power minimization. This research employed AI with the classic Traveling Salesman Problem (TSP) to provide a low-power synthesis. Initially, from an input State Transition Graph (STG), a Markov process probabilistic model is formed and the next KL algorithm is applied using the partitioning method where it calculates the Weighted Hamming Distance (WHD). To find the minimum power with respect to proper state encoding using SIS tool is a NP-Hard problem and rather time consuming. When the database has been generated, a Random-Forest regression model is trained to replace the SIS tool in order to predict the power for multi-level realizations based on the database. The model is even more effective at making predictions, with a maximum test error of 4.584% and an average test error of 0.799 %. According to the results, the augmented-based framework accelerates the process on average 1.5 times faster than the traditional simulation-based framework.