【摘要】 提出了一种复三层前馈网络的新型学习算法。该算法采用的是分层优化方法 ,将隐层的非线性神经元线性化 ,线性化产生的误差通过罚项受到限制。分层优化使得每一层权值整体优化 ,而与另一层无关 ,这样使得整个优化过程更有效。更多还原

【Abstract】 In this paper, a new complex algorithm for training of 3-layered perceptrons is proposed, which is based on a linearization of nonlinear processing neurons and the optimization of multilayer perceptrons layer by layer. The weights in each layer are updated depending on each other in one iteration epoch, but separately from other layers.\;更多还原