Effect of templateSV​M('Standar​dize',true​);

Dear all,

Is there any documentation that explains the advantage of using

t = templateSVM('Standardize',true)

when using

classifier = fitcecoc(featuresTrain,YTrain,'Learners',t);

Because in my case it provides a better classification result, but need to understand some basics on how it works.

Any comment is appreciated.

Meshoo

[Joy Li]

I have the exact problem. It seems to provide better classification accuracy compare to amy own settings. But I am not sure what templateSVM('Standardize',true) does.

[c]

%% template = templateSVM('Standardize',true)

% 'BoxConstraint' — 1 (default)

% 'CacheSize' — 1000 (default)

% 'ClipAlphas' — true (default)

% 'DeltaGradientTolerance' — 0 if the solver is ISDA (for example, you set 'Solver','ISDA')

% 'GapTolerance' -0 (default)

% 'IterationLimit' — 1e6 (default)

% 'KernelFunction' — 'linear' Linear kernel, default for two-class learning

% 'KernelOffset' -0.1 if the solver is ISDA (that is, you set 'Solver','ISDA')

% 'KernelScale' — 1 (default)

% 'KKTTolerance' — Karush-Kuhn-Tucker complementarity conditions violation tolerance

% 1e-3 if the solver is ISDA (for example, you set 'Solver','ISDA')

% 'NumPrint' — 1000 (default)

% 'OutlierFraction' —0 (default)

% 'SaveSupportVectors' —true (default) Store support vectors, their labels, and the estimated α coefficients

% 'ShrinkagePeriod' — 0 (default)

% 'Solver' — The default value is 'ISDA'

% 'Standardize' — false (default)

% 'Verbose' — 0 (default)

[c]

template = templateSVM('Standardize',true)-->

it means using all the default values.