Scribd
Upload
1 views

Kernel-Functions-in-Support-Vector-Machines

kernel fn

Uploaded by

kanhaiyamittal8755
Copyright
© © All Rights Reserved
Available Formats
Download as PPTX, PDF, TXT or read online on Scribd
1 views

Kernel-Functions-in-Support-Vector-Machines

kernel fn

Uploaded by

kanhaiyamittal8755
Copyright
© © All Rights Reserved
Available Formats
Download as PPTX, PDF, TXT or read online on Scribd
You are on page 1/ 10

Kernel Functions in

Support Vector
Machines

Kernel functions are crucial for enabling Support Vector Machines


(SVMs) to tackle complex, non-linear classification problems. This
presentation explores 12 commonly used kernel functions and their
applications in transforming data to higher dimensions.
Exploring Kernel Functions
Linear Kernel Polynomial Radial Basis Sigmoid Kernel
Kernel Function (RBF)
Kernel
The simplest kernel, Introduces polynomial Similar to the neural
performing linear transformations, creating Highly popular, mapping network's sigmoid
separation. Effective for higher-order decision data to infinite- activation function,
linearly separable data. boundaries. Useful for dimensional space. creating non-linear
capturing non-linear Creates non-linear boundaries. Used for
relationships. boundaries, suited for problems with
complex data patterns. overlapping classes.
Linear Kernel
K(x,y) = xTy

The linear kernel calculates the dot product between two data
points (x and y). It is the most basic kernel, suitable for linearly
separable data. However, it lacks the flexibility to handle complex
non-linear relationships.
Polynomial Kernel

K(x,y) = (gamma * xTy + c)d

The polynomial kernel introduces a polynomial transformation to


the data. The degree 'd' controls the complexity of the decision
boundary. Higher degrees can lead to overfitting if not chosen
carefully. 'gamma' and 'c' are parameters that influence the shape
of the boundary. It's useful for capturing non-linear relationships in
data.
Radial Basis Function
(RBF) Kernel

K(x,y) = exp(-gamma ||x - y||2)

The RBF kernel uses a Gaussian function to calculate the similarity


between two data points. It is a powerful kernel, capable of
mapping data to an infinite-dimensional space. The 'gamma'
parameter controls the influence of the Gaussian function,
impacting the smoothness of the decision boundary.
Sigmoid Kernel
K(x,y) = tanh(gamma * xTy + c)

The sigmoid kernel utilizes the hyperbolic tangent function (tanh)


to map data non-linearly. It is inspired by the sigmoid activation
function used in neural networks. 'gamma' and 'c' parameters
influence the shape of the sigmoid function and the resulting
decision boundary.
Laplacian Kernel
K(x,y) = exp(-gamma ||x - y||)

The Laplacian kernel uses the exponential function to measure the


similarity between data points. It is similar to the RBF kernel but
has a different influence on the smoothness of the decision
boundary. 'gamma' influences the rate of decay of the exponential
function.
Exponential Kernel

K(x,y) = exp(-gamma ||x - y||)

The exponential kernel is another kernel that employs the


exponential function for similarity calculations. It is used for
problems where the data points are not linearly separable.
'gamma' affects the rate of decay of the exponential function,
impacting the decision boundary.
Hyperbolic Tangent
Kernel
K(x,y) = tanh(gamma * xTy + c)

The hyperbolic tangent kernel employs the hyperbolic tangent


function (tanh) to perform non-linear mapping. It is similar to the
sigmoid kernel but has a different influence on the shape of the
decision boundary. 'gamma' and 'c' parameters control the shape
of the tanh function, impacting the decision boundary.
Rational Quadratic
Kernel

K(x,y) = (1 + gamma ||x - y||2)-1

The Rational Quadratic Kernel uses a rational function to calculate


similarity. It is often used for problems where the data has varying
levels of noise. 'gamma' parameter controls the influence of the
rational function, impacting the shape of the decision boundary.

You might also like