Hard and Soft Computing,
There are many tools in mathematics, but all of them can divide into Hard and Soft two kinds. Such as
A) Traditional physical is focus on Hard Computing, whereas
B) In 17th there are emerge statistic in mathematics, present, a new paradigm break out, it is fuzziness.
Why will emerge randomness (statistic) and fuzziness? The reason is Complex Problem. In Classical Mechanics, we can easy modeling a object move. However when we want to calculate a coin weather a head or tail, there are a great many of constraint to effect it. Indeed, we can not clearly understand the overall constraint, thus, we propose a new approach that statistic and get an approximate answer.
Hard Computing against Soft Computing,
Hard and Soft Computing is two different attitude of science. Many physical scientists are not support using soft computing in scientific problem solving. If a human is a computer, which will be a soft computer rather than a hard computer, there are many reasonable instances, such as we read a book, remember the entire story is not necessary, and just keep the key point is enough. In soft computing, it has a very powerful at dealing with complex problem, with Granular and Fuzzy mechanism. In common sense, we only finish the job, no matter how accuracy or precise it is. Thus, in many our real word problem, we just want to have an approximate answer.
From Hard to Soft Computing,
There is because of we have knowledge, that we can solve the problem. In our common sense knowledge, there is much knowledge with imprecise, incomplete, and not totally reliable. Soft Computing have many application, I remember that GuoLiang Chen academician visit our school, and he said that, he exploit Artificial Neural Network (ANN) in dealing with the high dimension and complex relation with learning character of ANN.
Many systems have not learning mechanism, whereas in Knowledge-Based System, learning is a very useful factor in application domain. Such as, you are a boss in a store. you want to have a function which can extract knowledge (rules) form large data set. That you need the learning function, in this requirement, ANN is useless, and should make use of fuzzy rule-based system.
Conclusion,
There are also many reasons why we should shift from hard computing to soft computing. The Machine Leaning is the most vivid instance in application. And it have widely use in many industry, such as bank, business support, finger recognize and so on. In words, Soft Computing is a good tools.