论文部分内容阅读
提出了一种用于多层前馈神经网络训练的新算法,它把遗传算法与自适应共轭梯度学习算法集成起来。这种并行混合学习算法已经在多指令流多数据流(MIMD)平台实现。通过把该算法用于一个图像识别问题,对它的性能进行了评估。文中还显示了所提出的并行混合神经网络学习算法良好的收敛性。
A new algorithm for multi-layer feedforward neural network training is proposed, which integrates genetic algorithm and adaptive conjugate gradient learning algorithm. This parallel hybrid learning algorithm has been implemented on the multi-instruction flow multi-stream (MIMD) platform. Its performance is evaluated by using the algorithm for an image recognition problem. The paper also shows the good convergence of the proposed parallel hybrid neural network learning algorithm.