Machine learning, data mining, and other

Machine learning, data mining, and other

In this book, we constantly mention "intelligence". What is "intelligence "? Are we talking about artificial intelligence? Or machine learning? What does it have to do with Data Mining and soft computing? In academia, the exact definition of the content introduced in this book has been controversial for several years. From the perspective of practice, there is no substantial difference between these concepts, more of which refer to the differences in the application environment. This book integrates the essence of all the above fields. Let's take a look at it one by one.

Artificial intelligence, which is widely known as AI, originated in the computing field in 1950s. Initially, AI was ambitious to develop machines that could think like humans (Russell and norvig, 2002; Buchanan, 2005 ). Over time, the target becomes more feasible and specific. The unattainable goal has to succumb to the cruel reality, but many of the fields we mentioned previously come from artificial intelligence, such as machine learning, data mining, and soft computing.

Now, even the most advanced computing intelligence system cannot understand the story that a 4-year-old reads. So, if we cannot let computers "think", can they "Learn? Can computers be taught to determine species based on animal characteristics? What about identifying poor secondary mortgages? More complex things, such as speech recognition and natural language reply, Can computers do it? The answer to all these questions is yes. However, you may be curious, "What are these problems ?". To solve these problems, the simplest way is to build a large data table in the computer and store the answers to all possible questions, then, you only need to search for existing answers in the table to answer the question.

Of course, the Data Table query method is feasible, but there are also some problems. First, in the actual product system, the table containing all the questions and answers must be very large. Therefore, from the perspective of efficiency, this is definitely not an optimal solution. Second, if there is no answer to a question in the database, you will not be able to give an answer. If a real user asks these questions, you can only use "sensitive words" to intercept them. Finally, people must be arranged to build and maintain the query table. As the table grows, the number of people required will also grow, which may make the company's financial department angry. Therefore, querying a table is not a good solution. We need a better solution.

Machine Learning refers to the ability of software systems to abstract general rules from existing experiences and then use these rules to answer various questions, including those that have been encountered and never seen. Some algorithms are transparent to humans, meaning that humans can understand the rules abstracted by algorithms. Typical examples of transparent algorithms include decision trees and all rule-based learning methods. Another type of algorithms is not transparent to humans. For example, neural networks and SVM belong to this type of algorithms.

Remember that, like human intelligence, machine intelligence is also unreliable. In the field of intelligent application, you will learn how to deal with uncertainty and ambiguity. Just like in the real world, the answers to all questions have a credibility, not an absolute reliability. Although in our daily life, we always simply assume that something will happen. Because of this, when using smart applications, we need to solve problems such as credibility, effectiveness, and the cost of errors.

 

This article is excerpted from intelligent web algorithms.

Book details: http://blog.csdn.net/broadview2006/article/details/6702401