ai决策_人工智能中的不确定性决策

ai决策

When the conditions are not certain for a computer system or an Artificially Intelligent agent, then there needs to be a method through which the agent can take decisions in uncertain situations also. For this, many methods were proposed by different developers and computer professionals. The following three are among the basic and widely used methods to perform decision making while handling the uncertainty:

当计算机系统或人工智能代理的条件不确定时，则需要一种方法，通过该方法，代理也可以在不确定的情况下做出决策。 为此，不同的开发人员和计算机专业人员提出了许多方法。 以下三种是在处理不确定性时进行决策的基本方法和广泛使用的方法：

1. Probabilistic Theory

   概率论

2. Fuzzy Logic

   模糊逻辑

3. Truth Maintenance

   真相维护

1)概率论 (1) Probability Theory)

The main reason for proposing this theory for reasoning was that there are situations where we cannot decide whether our statement or event is completely true or completely false. In most of the cases, the conditions appear to be partially true. So, this representation was not possible in the Boolean knowledge representation. So, to overcome this problem, the Probabilistic theory was introduced in which we can determine how likely an event is to happen. This estimation is done by assigning a real number within the range 0 to 1 where 0 represents that the event will never occur during the whole experiment, whereas 1 represents that whenever the experiment will be performed, that event will surely take place.

提出这种推理理论的主要原因是，在某些情况下，我们无法决定我们的陈述或事件是完全正确还是完全错误。 在大多数情况下，条件似乎部分正确。 因此，在布尔知识表示中不可能实现这种表示。 因此，为了克服这个问题，引入了概率论，其中我们可以确定事件发生的可能性。 通过在0到1范围内分配一个实数来完成此估算，其中0表示该事件在整个实验过程中永远不会发生，而1表示无论何时进行该实验，该事件肯定都会发生。

2)模糊逻辑 (2) Fuzzy Logic)

Fuzzy Logic is also somewhat similar to Probability theory. The difference is that, in fuzzy logic, the agent is trained to imitate the reasoning process of humans. Fuzzy Logic is termed better than Probabilistic theory in Artificial Intelligence because it emphasizes the system to think, analyze and act humanely. In fuzzy logic also, all the possible values between the range 0 to 1 are considered and the system produces a real number lying between this range which determines the possibility of the event.

模糊逻辑在某种程度上也类似于概率论。 区别在于，在模糊逻辑中，对代理进行了训练以模仿人类的推理过程。 模糊逻辑在人工智能方面比概率论更好，因为它强调系统进行人性化的思考，分析和行动。 同样在模糊逻辑中，考虑范围在0到1之间的所有可能值，并且系统会产生一个介于该范围之间的实数，该实数确定发生事件的可能性。

3)真相维护 (3) Truth Maintenance)

The Truth Maintenance process of decision making works on the basic logic that the knowledge that a system considers valid till now can become invalid in the situations which may arise in the future. This system is very useful to deal with the real world that it solves the most basic problem of uncertainty, which is the dynamic environment. As the system's surroundings keep changing as the time passes, the Truth Maintenance System keeps a record of the data which is valid till now. The data which becomes invalid is either moved to trash or is stored in some other file. The reason behind storing this data in some other file is that there may occur situations in the future where this data can again turn out to be valid.

决策的“真理维护”过程基于以下基本逻辑：系统认为到现在为止有效的知识在将来可能出现的情况下可能变得无效。 该系统对于解决现实世界中最基本的不确定性问题(即动态环境)非常有用。 随着时间的流逝，系统的周围环境不断变化，Truth Maintenance System会保留迄今为止有效的数据记录。 无效的数据将被移至废纸or或存储在其他文件中。 将该数据存储在其他文件中的原因是，将来可能会出现这种情况再次变为有效的情况。

翻译自: https://www.includehelp.com/ml-ai/decision-making-with-uncertainty-in-artificial-intelligence.aspx

ai决策