多Agent自动协商中机器学习的应用研究

目前将机器学习理论应用到多Agent自动协商系统中已成为电子商务领域的最新研究课题。本文即是利用贝叶斯法则来更新协商中的环境信息（即信念），利用强化学习中的Q学习算法生成协商中的提议，建立了一个具有学习机制的多Agent自动协商模型。并且封传统Q学习算法追行了扩充，设计了基于Agent的当前信念和最近探索盈余的动态Q学习算法。实验验证了算法的收敛性。

Research on Applying Machine Learning to Automated Negotiation in Multi-agent System

At present applying machine learning to automated negotiation in multi-agent system becomes the hotspot research in the field of eleclronic commerce. In this paper, we use Bayesian learning to revise beliefs, and put Q-learning algorithm to propose counteroffers of negotiation, and we establish an automated negotiation lnodal with learning mechanism. At the same time, we extend the traditional Q-learning into a dynamic Q-learning algorithm by introducing current beliefs and recent exploration bonus, the results of experiment show that our algorithm is convergent.