mlatoz

Kernel SVM

Mapping to a Higher Dimension

Mapping to a Higher Dimensional Space can be highly compute-intensive. So, it may require a lot of computation, a lot of processing power, and, you know, the larger your data set, the more of a problem this can cause.

The Kernel Trick

The Gaussian RBF Kernel:-

K(x-bar, lⁱ-bar) = e^{-(||x-bar - lⁱ-bar||² / 2σ²)}

  where,
      K = A Kernel that is a function applied to two vectors.
      x-bar = Point in a dataset
      l = Landmark
      i = No. of landmarks

Types of Kernel Functions

Gaussian RBF Kernel: K(x-bar, lⁱ-bar) = e^{-(||x-bar - lⁱ-bar||² / 2σ²)}
Sigmoid Kernel: K(X, Y) = tanh(γ.X^TY + r)
Polynomial Kernel: K(X, Y) = (γ.X^TY + r)^d, γ > 0

Non-Linear Kernel SVR (Advanced)

Heads-up about Non-Linear SVR

Section on SVR:
- SVR Intuition
Section on SVM:
- SVM Intuition
Section on Kernel SVM:
- Kernel SVM Intuition
- Mapping to a higher dimension
- The Kernel Trick
- Types of Kernel Functions
- Non-Linear Kernel SVR

Download Resources