基于神经网络的协商学习机制

多agent协商研究中,如何通过学习提高协商效率是一个重要的课题,目前的研究多采用简单的学习算法,学习效果不好。论文首先提出了一个两方多回合交互协商框架,然后依据协商历史结果、协商双方初次出价等信知,对协商结果信息进行预测,从而确定协商交互中的推理策略,并利用BP神经网络的自适应、自学习能力对协商结果预测机制进行学习。随后的验证系统表明,该机制通过对协商结果的有效预测,提高了协商交互的效率和协商个体的效用。

A Negotiation Learning Machine Based on BP Network

In MAS System,how to learn in negotiation is one of the most important part,but till now researcher focus only on some simple learning method.This paper proposes a bilateral multi-round negotiation model.Then,based on history result of the negotiation and the first proposal of both negotiators,a machine has been proposed to forecasting the result-set of the negotiation,thus to determine the reasoning tactic in negotiation.In addition,we have proposed a learning machine based on BP neural network in order to optimize the forecasting mechanism.The following experimentation show that this machine can maximize the negotiation utility of participants and improve the negotiation efficiency.