Why there is no Quadratic kernel in fitcsvm

Hi;

In SVMTrain there is Quadratic kernel function. However, I notice in he new fitcsvm . There is no Quadratic ... any reason for that?

Thanks!

Answer 1

[Krishna]

Hello Diver,

From what I gather, you seem to be encountering difficulty locating the 'Quadratic' kernel option within the ‘fitcsvm’ function. The ‘fitcsvm’ function in MATLAB is designed for training a binary classification support vector machine (SVM). It offers a variety of kernel functions, such as linear, polynomial, Gaussian (also known as radial basis function or RBF), and sigmoid. If you're accustomed to older MATLAB versions or different SVM libraries, you might recall the explicit "quadratic" kernel, which is simply a polynomial kernel with a degree set to 2.

To use a quadratic kernel in the fitcsvm function, you would set the 'KernelFunction' parameter to 'polynomial' and then specify the 'PolynomialOrder' parameter as 2.

Please go through this documentation to learn more,

https://www.mathworks.com/help/stats/fitcsvm.html#namevaluepairarguments

I hope this clarifies your issue.