An improved Gaussian dynamic particle swarm optimization (PSO) algorithm is proposed in this paper. In the proposed version of PSO, the original swarm of particles is initialized by canonical PSO. The time varying linear inertial weight is reintroduced to add to the position update formula. And the craziness variable is also used in order to maintain the diversity of particle swarms. The performance of improved Gaussian dynamic PSO is demonstrated by applying it to several benchmark functions and comparing to other variants of PSO.