A Novel Hybrid Analog Design Optimizer with Particle Swarm Optimization and modern Deep Neural Networks

This work presents a novel hybrid optimization technique that combines a Particle Swarm Optimization (PSO) engine with a multi-output Deep Neural Network (DNN) to obtain a fast and accurate analog circuit optimizer. A Deep Learning supervised regression model is used to replace the slow simulations required in the standard PSO. A CMOS miller-opamp is used as the design problem. Using the hybrid PSO-DNN technique has combined the speed of the DNN model and the accuracy of the PSO. Moreover, Deep Learning modeling has improved the accuracy compared to the standard machine learning techniques.