the dynamic behaviors of continuous neural networks under structural variations in learning process are studied. A mathematical model of continuous neural networks specified by a time variation nonlinear differential equation is presented. A linearization method for analysing dynamics of a nonlinear network with structural perturbations at equilibria is suggested. The exponential stability of associative neural networks with invariant structure and with bounded slowly variation structure are investigated.