基于人工神经网络的负相关学习研究

人工神经网络集成技术是神经计算技术的一个研究热点,在许多领域中已经有了成熟的应用.神经网络集成是用有限个神经网络对同一个问题进行学习,集成在某输入示例下的输出由构成集成的各神经网络在该示例下的输出共同决定.负相关学习法是一种神经网络集成的训练方法,它鼓励集成中的不同个体网络学习训练集的不同部分,以使整个集成能更好地学习整个训练数据.改进的负相关学习法是在误差函数中使用一个带冲量的BP算法,给合了原始负相关学习法和带冲量的BP算法的优点,使改进的算法成为泛化能力强、学习速度快的批量学习算法.

Negative Correlation Learning Based on Neural network

Artificial Neural network ensemble (ANNE) is a hot topic of neural computing, which has been maturely applied in many fields. A neural network ensemble is a very successful technique where the outputs of a set of separately trained neural network are combined to form an unified prediction. Negative correlation learning (NCL) algorithm for training ANNE is introduced in this paper to encourage different individual networks in an ensemble to learn different parts or aspects of a training data so that the ensemble can learn the whole training data better. NCL in this paper can create negatively correlated neural networks using a correlation penalty term in the error function. It can also combine the advantages of original negative correlation learning and BP algorithm with impulse, thus making the developed algorithm into a batch algorithm with competed extending ability and fast learning speed.