Efficient rate adjustment hardware for on-chip learning
M.G. Rezaie, Farzan Farbiz, Ashkan Behnam · 2005
This paper presents a new approach to facilitate the implementation of adaptive adjustment of on-chip rate learning in feed forward neural networks. A typical multi layer perceptron (MLP) network with controlled learning method is assumed as the target of the design and a Gilbert amplifier cell is used to tune an adaptive learning rate to find an optimum trajectory for robust design.