You should have a good understanding of the classification models that are and will continue to be popular. Random forest and Support Vector Machines (SVM) Random forest was briefly introduced in the previous article, and this article will briefly introduce support vector machines (SVM). Emphasis on application, light on mathematical interpretation and derivation.
Linear classifier
A very simple classification problem.
But of the myriad possible lines, which works best?
In SVM, it becomes Maximum Marginal
Machine learning algorithm (2)- Support vector machine (SVM) basis
Linear indivisibility
Since there are too few cases of linear indivisibility, the following diagram is a typical classification diagram of linear indivisibility.
- Separate them completely with a graph
- The other one is still a straight line, which doesn’t guarantee separability, tolerates errors.
In the second case, if the penalty function makes the misclassification of phi as reasonable as possible. A penalty can be added to a misdivided point, and the penalty function for a misdivided point is the distance from the point to its correct position:
Kernel function
As I mentioned, you can use nonlinear methods to partition perfectly. Let’s go from a linear space to a higher dimensional space, and in this higher dimensional linear space, we’re dividing it by a hyperplane.
The above is a simple understanding of SVM, which the author does not quite understand. I just make a record and look forward to having the ability to fill in the holes later.