A model of cellular neural network with transient chaos is proposed, in which a negative self-feedback is
introduced into a cellular neural network after transforming the dynamic equation to discrete time via Euler's method. The simulation of single neuron models shows the characteristics of bifurcation and chaos. In optimization, the model gradually approaches, through a chaos search by the course of reversed period-doubling bifurcations, to a dynamical structure similar to the Hopfield neural network which converges to a stable equilibrium point. As the model has rich dynamics such as randomicity, it can be expected to have robust search ability for global optimal solutions. Finally, simulation results on two examples of function optimization show the effectiveness of the neural networks.