• Support Vector Machine (SVM) is a supervised Machine Learning algorithm.
• SVM is particularly well suited for classification of complex but a small or medium sized data set.
• SVM can be used for classification (SVC), regression (SVR), and outliers detection.

What are data points, or samples in SVM?

Dataset tagged as + and -.

What is a hyperplane?

A N-1 dimensional subspace for an N dimensional space, such as a line in x-y coordinate, or a plane in 3d world. Wikipedia.

The red curve and straight lines are both hyperplane for a 2d dataset.

What are Functional Margins?

-1 →| 0 |← +1

↖ Dotted lines

Like streets; the widest line around hyperplane that has support vectors on it.

What are Support Vectors?

.

↖ Black and Grey dots on dotted line.

Data points lying closest to the hyperplane and the most difficult to classify.

"Off-street" samples are not SV.

What is Support Vector Machine?

• Supervised learning model.
• Analyze data for classification and regression.
• Can perform both linear & non-linear classifications.

What is Linearly Separable?

• A proper of two sets of points.
• Visualized as a line in the plane separate two sets of points.
• w: n + 1 real numbers; X: n-d vector; For set 1, exist k such that:

$$ w^T X < k $$

What is Soft Margin?

Same samples are allowed to be between margins.

What is Hard Margin?

All samples must be off the street.

What is Hinge Loss?

↘️→

max(0, 1-t).

What is the idea of SVM Regression?

Fit as many samples on the street.

What is the idea of SVM Clustering?

paper

• Data points are mapped to a high dimensional space (Gaussian kernel).
• Search for the minimal enclosing sphere.
• Map such a sphere back to data space.

What is the motivation of SVM?