On the training of neural network for associative memory

In this paper, an optimized training scheme of neural network for associative memory was proposed. It shows that the basins of attraction for samples attraction can be controlled in some extent by a pitfall depth parameter. Therefore, the fault-tolerance of network can be made as good as possible. Numerical simulations show that using this scheme, the capacity of network can reach ��?1 (��=M/N, here N is the number of neurons and M is the number of stored sample) and still with good fault-tolerance. The results are much better than the popular schemes such as outer-product scheme, orthogonalized outer-product scheme, pseudo-inverse matrix scheme, etc.. The problems on symmetry and convergence of trained networks were discussed too.