推测计算中多Agent的资源协商

推测计算是在信息不完全的情况下,提前进行决策的一种计算过程.它以假设推理为基础,并利用缺省值对未来的问题进行计算.事实上,在计算过程中,缺省值并不是唯一的基础信息.在时间允许的情况下,主Agent可以通过协商获得尽可能多的真实信息,信息越丰富,推测计算的结果越精确.为了强化协商,本文分析了假设推理和推测计算的基本理论,并以此为基础给出了推测计算的扩展框架,定义了协商语言,提出了协商算法,并将其嵌入到了推测计算中,从而可以提高推测计算的精确性.

Multi-Agent Resource Negotiation in Speculative Computation

The speculative computation is a kind of computing process that is used in decision-making in advance under incomplete information. It is based on abduction and utilizes defaults to compute the future problem. In fact, in computing process, the defaults are not the only basic information. Under situation that time allows, master agent can obtain real information as more as possible via negotiation. The more information is, the more accurate the result is. To strengthen negotiation, basic theories of abduction and speculative computation are introduced, based on these theories, framework extended of speculative computation is presented, and negotiation language and negotiation algorithm are presented, and the algorithm is embedded in speculative computation. Thus it can improve the accuracy of speculative computation.