人工智能对不确定性的适应性表征

Adaptive Representation of Artificial Intelligence to Uncertainty

自然世界是复杂的和不确定的,它并不像决定论所描述的那样是确定有序的,严格遵循自然规律。在表征的意义上,决定论的世界相对较易表征,非决定论世界中的不确定性现象或事件如何表征却是我们面临的一个难题。在人工智能中,不确定性往往表现为统计性和概率性,概率表征就自然成为一种常见方式,如贝叶斯网表征。可能世界作为一种不确定性的表达,可由相关概率模型、基于逻辑规则方法、D-S方法和模糊集-模糊逻辑方法表征。除了这些静态表征方式,不确定性也可通过动力学概率表征。

The natural world is complex and uncertain.It is not deterministic and orderly,as determinism has described that it strictly follows the laws of nature.In the sense of representation,the deterministic world is relatively easy to represent,but how to represent the uncertain phenomena or events in the non-deterministic world is a difficult problem for us.In artificial intelligence,uncertainty is often showed as statistics and probability,so probable representation naturally becomes a common way,such as Bayesian network representation.Possible worlds,as an expression of uncertainty,can be represented by related probability model,logic-based method,D-S method,and fuzzy set-fuzzy logic method.In addition to these static methods of representation,uncertainty can also be represented by dynamic probability.