Original link: https://github.com/jackfrued/Python-100-Days
[Python - 100 days since the novice to master]
Machine learning foundation
The so-called "machine learning" is the use of computer complicated data into useful information, so that you can explore the significance of the data to bring law and hide behind the data. Now, "machine learning" and "big data" can be said in two words hottest IT industry, the nature of the problem and whether it is "machine learning" or "big data" is ultimately resolved is the same, with the most straightforward words is to use existing data to predict future conditions.
According to the "input" and "output" problem, we can solve the problem with a computer is divided into four categories:
- Information input to be accurate, the optimal solution required output.
- Enter the information is accurate, we can not find the optimal solution.
- The information input is ambiguous, the optimal solution required output.
- Enter the information is vague, unable to find the optimal solution.
In the four major categories of issues above, the problem is the computer class 1 best at solving of such issues is actually a problem "numerical" and "logical reasoning" aspects, and artificial intelligence in the traditional sense is the use of logical reasoning to solve the problem (such as the early "man-machine chess"). All along, we are accustomed to the computer known as "computer", and based on the "von Neumann" architecture "PC" is really just realized the function "human brain" rational thinking part of this, but at this point "computer" is usually better than the "brain", and the "brain" manifested in dealing with fuzzy information input processing power, today it seems not a "computer" can fully match. So we study artificial intelligence Ye Hao, Ye Hao study machine learning, is to enter fuzzy information, the computer can give a satisfactory or even the best answer.
At this point, we can give "machine learning" under a definition: machine learning is how a specialized computer simulation or realization of human learning behavior to acquire new knowledge or skills, re-organize existing knowledge structures so as to continuously improve their disciplinary performance. Machine learning has been widely applied to various areas of production and life, Here are some classic scenes:
- Search engine: based on your search habits, Optimization of the results of a search.
- Electricity supplier website: automatic Other products that might interest you.
- Loan application: a comprehensive assessment by your recent financial events.
- Image Recognition: Automatically recognizes the picture there is no discord content.
Machine learning can be divided into supervised learning and unsupervised learning. Supervised learning function is obtained from a given training data set to learn when new data arrives, this function can predict the results, supervised learning the training set consists of input and output, it can be said characteristics and objectives. Supervised learning target is marked by people, rather than supervised learning there is no data type information, training set and no artificial result marked by unsupervised learning can reduce the dimension data feature, so that we can use more two-dimensional or three-dimensional graphics intuitive display data.
The general steps to implement machine-learning:
- data collection
- data preparation
- data analysis
- Training algorithm
- Test algorithm
- Algorithm