人工智能和神经网络笔记8.ppt

“备”则“倍” 有准备、有规划的人生更精彩！ i.e., U is exactly a stable state of the network, and W thus determined is reasonable. Example Given n=3, m=3 U =(1 1 -1), U =(-1 1 1), U =(1, -1 1) Thus 1 2 3 t t t U U -I = U U -I = U U -I = W = ? (U U -I) = 1 1 t 0 1 -1 1 0 -1 -1 -1 0 2 2 t 3 3 t 0 -1 -1 -1 0 1 -1 1 0 0 -1 1 -1 0 -1 1 -1 0 k k t k=1 3 0 -1 -1 -1 0 -1 -1 -1 0 WU = 1 0 0 -2 Sgn(WU ) = 1 1 -1 = U 1 1 Similarly WU = 0 0 -2 Sgn(WU ) = 2 2 2 -1 1 1 = U WU = 0 -2 0 Sgn(WU ) = 1 -1 1 = U 3 3 3 Clearly, U , U , and U are stable memories. The structure of the network is as below: u u u 3 2 1 -1 -1 -1 Applications: a) Classification Given input U = (1 1 1), whether U ? {U , U , U }? x x 1 2 3 WU = x -2 -2 -2 Sgn(WU ) = -1 -1 -1 ? U x Hence U does not belong to the set {U , U , U }. 1 x 2 3 b) Associative Memory Given a noisy input U = (0 1 -1) , what is U ? t x t x WU = x 0 1 -1 Sgn(WU ) = x 1 1 -1 = U . 1 U ? U . x 1 In general, if n-dimensional vector U , …, U are orthogonal, nm, then 1 m WU = (U U - I)U + ? (U U - I)U k k k k k t t i i m i=1 i?k = U (U U - IU ) + ? (U U U - IU ) k k k k k k t t m i=1 i i i?k = U n - U + (m-1)(-IU ) = (n-1)U - (m-1)U = (n-m)U k k k k k k Sgn(WU ) = Sgn[(n-m)U ] = U , k = 1, 2, …, m k k k Hence {U } are stable states. k * * AI NN Notes Chapter 9 Feedback Neural Networks §9.1 Basic Concepts Attractor: a state toward which the system evolves in time starting from certain initial conditions. Basin of attraction: the set of initial conditions which initiates the evolution terminating in the attractor. Fixed point: if an attractor is in a form of a unique point in state space. Limit Cycle: if an attractor consists of a periodic sequence of states. Hopfield Network and its Basic Assumptions 1 2 n v v v 1 2 n T T T 1 2 n i i i n 2 1 w w w w 21 12 n1 w w n2 1n 2n 1. 1 layer, n neurons 2. T -- T