贝叶斯因果网络_因果关系和贝叶斯网络

贝叶斯因果网络

Thinking causality is necessary to act, therefore anticipating, prognosticating, designing, supervising, evaluating, diagnosing organizations. The question that arises is: what is the nature of causality? We therefore have a universe of possibilities, and causality describes effect E, which necessarily occurs, all the time, when a certain cause C occurs. We therefore have a deterministic and universal approach to causality. This approach is challenged by the probabilistic approach and Bayesian logic.

要想因果关系就必须采取行动，因此要预测，预测，设计，监督，评估，诊断组织。 出现的问题是： 因果关系的本质是什么？ 因此，我们拥有各种可能性，并且因果关系描述了效应E ，当某种原因C发生时，效应E总是会发生。 因此，我们对因果关系采用确定性和通用的方法。 这种方法受到概率方法和贝叶斯逻辑的挑战。

概率方法： (The probabilistic approach:)

“Causes must be able to be characterized by this that they increase the probability of their effets” (Drouet, 2007). I do not know whether it is necessary for cause C to occur for effect E to necessarily occur, but I can say, with some caution related to my limited knowledge “a cause C increases the probability of its effet E in the sense that the conditional probability P(E|C) is higher than the absolute probability p(E)”

“ 因果关系必须具有这样的特征，即因果关系会增加其影响的可能性 ”(Drouet，2007年)。 我不知道是否必须使原因C发生才能使效果E必然发生，但我可以说，出于与我有限的知识有关的谨慎考虑，“ 原因C在一定条件下增加了其E的概率”概率P(E | C)高于绝对概率p(E )”

I can analyze a causal relationship without knowing all the causes, or assuming that even a finite set of all causes exists. We do not know of a set of factors to which the property of being a smoker would belong that suffirait to produce lung cancer. Above all, there is no guarantee that such a set exists”

我可以在不了解所有原因的情况下分析因果关系，也可以假设所有原因的有限集合都存在。 我们不知道吸烟的属性会导致肺癌的一系列因素。 最重要的是，不能保证存在这样的集合”

This does not mean that all causes are equal, but that they are conditional and that our knowledge about them is limited.

这并不意味着所有原因都是平等的，而是它们是有条件的，并且我们对它们的了解是有限的。

贝叶斯网络(BN)的特征 (Characteristics of Bayesian Networks (BN))

The originality of BN is to couple graph (causal) and probability. Bayesian network is composed of something other than the single oriented graph and a set of arrows constitutes a binary relationship on the set of variables that are vertices of the graph. In this post I propsoe a further explanaition:

BN的独创性是将图(因果关系)和概率耦合在一起。 贝叶斯网络由除单个定向图之外的其他东西组成，并且一组箭头在作为图顶点的变量集上构成了二进制关系。 在这篇文章中，我提出了进一步的解释：