On the state transition graph of Hopfield net model
C.Y. Lee, J.S. Wang, R.C.T. Lee · 1990
Relations of stable states between different Hopfield neural nets are shown, and upper bounds to their transient lengths are given. The proofs of the results lie in examining the sinks and the maximum path length of the state transition graph. IfWis the weight matrix of a Hopfield associative memory determined frommreference patterns, the experiments showed thatPmax(G(W,Θ)) is of order O(mn) andTord(GB(W,Θ))=2 whenm=1 or 2, for alln⩾3, wherenis the size of the net and Θ is zero factor