Linear SVM Classification

• Soft Margin Classification
• Hard Margin Classification

Soft margin allows for some outliers, so a few can be classified on the opposite side of the street.

C is the regularization hyperparameter.

Low value – very wide margins, underfitting.

Higher value – narrower margins.

If your margin is overfitting, try reducing C to regularize the model.

Hard Margin

Tries to find the line that absolutely separates the classifications.

Non Linear Classification

Not all datasets are linearly separable, so we have to add polynomial features, in some cases this can result in linearly separable features

• Polynomial Kernel

• Gaussian RBF Kernel

Polynomial Kernel

Low polynomial degree, it can't deal with very complex datasets.

If the model is overfitting, you may want to reduce the polynomial degree.

If the model is underfitting, then you may want to increase the polynomial degree.

No combinatorial explosion.