Hopfield网络的改进方法

Mokification of Hopfield networks

多重训练编码和傀儡编码使存贮在Ｈｏｐｆｉｅｌｄ网格中的模式即使不是局部最小点也能被回忆出来，某一特定模式可以通过多重训练编码得以回忆，并讨论了最少训练次数，通过附加傀儡数据，傀儡编码可确保回忆Ｈｏｐｆｉｅｌｄ网络中所有的存贮模式．

Multiple training enceding methed and puppet encoding methed guarantee recallingof stored patterns in Hopfield networks,even though they are not local minima.A single storedpattern can be recalled through multiple training encoding methed,and the minimal number oftimes for using the pattern for training to guarantee recalling of the pattern is derived.Withpuppet data,puppet encoding methed has guaranteed recalling of all stored patterns in Hopfieldnetworks.