This work gives an improved criterion for asymptotical stability of a class of neural networks described by multi-delay neutral differential equations. Introduced suitable Lyapunov-Krasovskii functional, a delay dependent criterion which not only depends on the discrete delays but also on the neutral delay is presented. The criterion,which can be solved by various efficient convex optimization algorithms, is expressed in terms of linear matrix inequality. In the end of the work, utilizing Matlab toolbox, the numerical example is presented to illustrate feasibility of the criterion given in the work.