Optimization of Decoupling Capacitors in VLSI Systems using Granularity Learning and Logistic Regression based PSO

In order to reduce power supply fluctuations and to maintain a low Power Delivery Network (PDN) ratio in high-speed Very Large Scale Integration (VLSI) systems, decoupling capacitors are used in power delivery networks. In order to lower the cumulative impedance of the PDN below the target impedance, an Adaptive Granularity Learning (AGL) and Logistic Regression (LR) based Particle Swarm Optimization (PSO) is used for optimization of decoupling capacitors, in this work. The proposed approach provides results very efficiently compared to the state-of-the-art approaches. A maximum gain of 81% in terms of CPU time is achieved compared to the conventional PSO based approaches.