一种快速的演化学习神经网络模型

神经网络的学习算法一直是人们研究的重点,各种算法都有自己的优点和不足。演化学习具有全局优化性能好,实验样本数要求少等优点,但也存在一些问题,如演化学习的速度较慢、模型的稳定性较差等。笔者在多年的神经网络的研究过程中,对传统的演化学习算法作了一些改进,该文利用梯度下降计算改进演化算法,极大地提高了模型的学习速度,对那些需要及时优化工艺参数的生产过程,该种模型有很好的应用前景。

A Fast Evolutional Learning Algorithm of Neural Network

The learning of neural network is researched zealously,each has its advantage.The evolutional neural network has a global optimum character with a few test.But it has some disadvantages,for example:the rate of evolutional learning is too slow sometimes,and the stability of network is discontented.In a deeply studying,the authors improve the evolutional algorithm.Specially,the rate of evolutional learning is increased by gradient descent algorithm,it is very useful to optimum the real-time craft process.