【摘要】 模式分类是RBF神经网络应用的一个重要方面。在线环境中数据集是经常变动的,采用批量式学习算法(如OLS算法)训练RBF网络会产生大量的重复训练,从而导致学习效率不高。为弥补这种不足,从梯度下降方法推导出一种增量式学习算法,用于在线环境中的RBF神经网络训练。最后将基此算法构建的在线分类系统用于IRIS分类问题。结果表明,该算法有较快的收敛速度,网络的在线分类性能良好。更多还原

【Abstract】 Pattern classification was an important part of the RBF neural network application. Under on-line environment, the training dataset was variable, so the batch learning algorithm(such as OLS) which would generate plenty of unnecessary retraining had a lower efficiency. It was deduced an incremental learning algorithm(ILA) from the gradient descend algorithm to improve the bottleneck. ILA could adaptively adjust parameters of RBF networks driven by minimizing the error cost, without any redundant retraining. Using the method proposed, an on-line classification system was constructed to resolve the IRIS classification problem. Experiment results showed the algorithm had fast convergence rate and excellent on-line classification performance.更多还原