On a neural network associative memory that uses indirect convergence

A special type of neural network associative memory which utilizes indirect convergence is introduced. During the synchronous, iterative recall process, every neuron state update must be in the right direction, i.e. no wandering transition is allowed. The tradeoff between the number of fundamental states and their attraction force is analyzed, under the constraint of probability of successful recall being not less than 0.99. The major advantage of such a network is its quickness in seeking for the stable state, even with a comparable number of stored states