I. BACKGROUND
In machine learning, there are 2 great ideas for supervised learning (supervised learning) and unsupervised learning (unsupervised learning)
Supervised learning, in layman's terms, is you know the answer to the question, need a computer to give a more standard answer.
Non-supervised learning, in layman's terms, is like a flock of birds. We get a lot of data, but don't know the answer to the question, hope that the computer to provide us with ideas.
In a production environment, mixed mode is often used. For example, image search, how to find the Web page to determine the picture is a tiger, that is a dog. There are 2 ideas.
1. According to the text around the picture.
2. Image data analysis of the picture.
2 angles to each other, stable, you can generate enough annotation information.
Two. Supervised learning
1. Supervised learning is the correct answer to a given dataset and data set, and the algorithm is based on this known dataset to make learning and predict results. Common regression and classification problems in supervised learning
1). Regression issues
For example: A number of known real estate data, each including the size of the house and price, to predict a given size of the house size and price, such problems we call regression problems.
2). Classification issues
For example: The medical community, based on the age of known cancer patients and the type of tumor (benign/malignant), wants to predict the type of tumor a patient is suffering from, which we call the classification problem.
2. Each sample has been marked as a positive or negative sample in the supervised learning data set
Three. Unsupervised learning 1. In unsupervised learning, the data we use will look a little different from what we watch in our study. There is no concept of attribute or label in unsupervised learning, which means that all data is the same without distinction.
So in unsupervised learning, we have only one dataset, and we don't know what each data point means. Instead, it only tells us that there is a dataset now, and you can find some kind of structure in it.
For a given dataset, unsupervised learning algorithms may determine that the dataset contains n different clustering.
2. Unsupervised learning algorithm, the most common is clustering. For example, Google News for the classification of news, divided into sports, hot, social ...
