1. Supervised Learning: We teach computers how to "learn", there are two big branches, one is regression, the other is classification. Supervised learning is a given "standard answer". Regression: "Standard answer" is continuous, according to the standard answer can make a curve function, you can estimate other unknown output data. For example, an estimate of three-month sales. Classification: "Standard answer" is discrete, can be divided into n types according to the current function (eg. split into two), enter the data, you can know which class, such as whether or not have breast cancer judgment. Supervised (classes is known to all):
K-nn
Np
Naive Bayes
Decision Trees
SVM unsupervised Learning there is no standard answer, clustering algorithm is a typical algorithm, no given function division, the essence of a similar type will be clustered together. For example, give you a bunch of data that lets you analyze the structure of this heap of data. Google News, for example, collects large amounts of news from major websites every day, and then gathers them all, automatically dividing them into dozens of different titles. A typical example of unsupervised learning is the cocktail party problem (separation of sounds), in which two sounds are received at the cocktail party, and two different microphones are picked up in different places, while unsupervised learning can be used to separate the two different sounds. Note that the reason for unsupervised learning is that there is no prior knowledge of the types of these sounds (the category here is the meaning of the label). and the code implementation of the subject as long as one line.
Yj
Links: https://www.zhihu.com/question/23194489/answer/75555668
Source: Know
Copyright belongs to the author. Commercial reprint please contact the author for authorization, non-commercial reprint please specify the source. Unsupervised (no Class):
K-means
Hierarchical clustering
[Coursa] Machine learning_1st Class
