一种基于Pareto最优概念和神经网络混合策略的多目标进化算法

基于Pareto最优的多目标进化算法得到了广泛地应用，但不适用于目标函数为非解析式的情况。基于神经网络和Pareto最优的联合策略，提出了一种解决此类问题的新方法：首先采用神经网络对历史数据进行学习，建立有效的神经网络模型来代替目标函数解析式；然后将神经网络模型嵌入多目标进化算法，进行进化计算；最后，将本文方法应用于卷烟配方比例掺配问题。实验结果表明，该方法优于传统方法，能较好地解决问题。

Multi-objective optimization algorithm with combined strategy based on Pareto optimal and neural network

Current multi-objective evolutionary algorithms cannot solve the problems of non-analytic arithmetic objective functions. With mixed strategy based on evolutionary algorithm and neural network, a novel method to deal with this kind of problem was proposed. Firstly, by training on historical data with neural network, the objective function could be replaced by feasible neural network model. Then, evolutionary computation could be followed by embedding the given neural network model in multi-objective evoluti...