A rigorous result about the off-line learning approximation

Summary form only given, as follows. The authors consider a mathematical justification of the offline approximation in continuous-time neural networks. In real-time models, a network behavior is characterized by two distinct dynamics evolving according to different time scales, the weight dynamics which is the 'slow' dynamics and the activation dynamics which is the 'fast' dynamics. The offline approximation consists in assuming that during the learning process neural activities are in their steady states. Such an approximation is a common dogma often used to provide an analysis of a network behavior. The authors considered convergent networks and proved that this approximation is valid on a large time scale on the order of 1/ epsilon where epsilon is the learning rate parameter which controls the learning velocity.<
>