Classification of the problem and how the computer solves these problems

The world's problems are divided into three types, one that has been found to be regular and can be obtained by applying finite-order laws to all solutions (optimal and other solutions). The second is to find the law, but its computational capacity is infinite. The third is that there is no regular problem.
The first problem is the simplest. In the early days of computer development, most of the problems were solved. such as scientific calculation, subtraction.
The second problem is divided into two categories. A class is a solution that has more than one reconciliation. Only one solution to the problem is very tricky, the only way is brute force. There are several solutions to the problem can be retired and second, through a certain optimization algorithm (greedy algorithm, what kind of tree used by the dog) to obtain suboptimal solutions.
The third is that there is no regular problem, which can be done by statistical means. For example, pattern recognition in AI. A system wants to recognize the face, but the world has yet to find an absolute formula to recognize the human face. But it can be solved indirectly by a statistical method. Through a large number of case training, constantly changing the variables factors, such as color division, shape distribution, contour, time (usually represented by vector) weights to balance the final result.
To make it clear here, the "law" I said above refers to the fact that the results can be directly expressed and calculated accurately by the data formula. Like the third problem is that there is no way to achieve 100% accuracy, statistics in itself is a kind of infinite approaching but not necessarily things.

Classification of the problem and how the computer solves these problems