LEARNING TIME OF LINEAR ASSOCIATIVE MEMORY

Neural networks can be used as associative memories which can learn pr oblems of acquiring input-output relations presented by examples. The learning time problem addresses how long it takes for a neural network to learn a given problem by a learning algorithm. As a solvable model to this problem we analyze the learning dynamics of the linear associ a tive memory with the least-mean-square algorithm. Our result shows t hat the learning time tau of the linear associative memory diverges in tau proportional to (1 - rho)(-2) as the memory rate rho approaches 1 . It also shows that the learning time exhibits the exponential depend ence on rho when rho is small.