Generating the optimized solution for obtaining a particular combination of design criteria is very affordable and time consuming. Optimization algorithms should carry out optimization process with high accuracy in a short time. An artificial intelligence algorithm that does not require complex equations and long simulation time for optimization of circuit topologies would be developed. Due to simple structure and fast response time, particle swarm intelligence algorithm is an ideal candidate as an optimizer. Algorithm would be trained with a data set consists of transistor dimensions and performance measurements, in other words algorithm would learn the circuit topology with all nonidealities and nonlinear occasions. Once trained, algorithm would not calculate complex circuit equations again and again; so faster response of optimization result would be obtained. © 2008 IEEE.