kernel scale in svm ?

In:

template = templateSVM(...
    'KernelFunction', 'gaussian', ...
    'KernelScale', 3.9, ...
    'BoxConstraint', 1, ...
    'Standardize', true);
modelSVM = fitcecoc(...
    predictors, respones, ...
    'Learners', template, ...
    'Coding', 'onevsone');

Kernel scale here: represents gamma or sigma ?

Answer 1

[Walter Roberson]

According to https://www.mathworks.com/matlabcentral/answers/516738-what-kernel-scale-in-svm-really-is#comment_983291

@Hiro Yoshino wrote

Kernelscale is literally a scaling parameter for the input data.

The input data is recommended to be scaled with respect to a feature before being applied to the Kernel function. When the absolute values of some features range widely or can be large, their inner product can be dominant in the Kernel calculation. So this kernelScale can be used to prevent this from happening. It also helps maintain the information.

[M. A.hsh]

The equation of the Gaussian (RBF) kernel can be expressed using either σ (sigma) or γ (gamma) depending on convention. In MATLAB’s fitcsvm, the parameter KernelScale corresponds to σ, or γ ?