[PYTHON] DeepRunning ~ Level 4.7 ~

Level4. Machine learning course (theory and practice)

What is a deep learning course that can be crushed in the field in 3 months

4-7. Support Vector Machine (SVM)

4-7-1. What is a support vector machine?

● Machine learning method for 2-class classification

● The distance between the linear discriminant function and the closest data point is called the “margin”. Find the linear discriminant function that maximizes the margin.

● Support vector The training data that constitutes the separation hyperplane is only the support vector, and the rest of the data is unnecessary.

● Soft margin SVM When the sample cannot be separated, the error is tolerated and the error is penalized. Even if linear separation is not possible, the decision boundary changes depending on the magnitude of parameter C.

● Non-linear separation When linear separation is not possible, it is mapped to the feature space and linearly separated in that space.

● Kernel trick The inner product of high-dimensional vectors is represented by a scalar. Calculation costs can be suppressed even if the feature section is high-dimensional.

• It is necessary to review the derivation of the objective function separately and cultivate the knowledge. (Future goals)

4-7-2. Hands-on

Let's run and check it based on the sample code.

4-7-3. Consideration

How to think about margins and how to visualize decision boundaries I felt that it would be easier for someone to understand when explaining. Support vector machines ... I find it difficult for people who don't understand. Tolerance of misclassification, proper relationships that give margins and penalties, etc. I'll give it a try.