MATLAB Answers

Why support vectors values are negative using regression learner app?

1 view (last 30 days)
Trained model from regression learner app (Gaussian regression) provides some negative value of support vectors. Why is it happening?

Accepted Answer

Aditya Patil
Aditya Patil on 29 Mar 2021
Support vectors are those data points that form the support vector, which are the two vectors (in higher dimension) which separate the data. So, if your data contains negative points, the support vectors may also be negative. Also note that preprocessing such as standardization changes input values, which can be seen from following example,
rng("default");
X = rand(100,1);
y = rand(100,1);
mdl = fitrsvm(X,y,'Standardize', true);
% These are the actual support vectors
supportVectors1 = mdl.SupportVectors;
% These are not the support vectors, but the data points which correspond to them
supportvectors2 = mdl.X(mdl.IsSupportVector);
sum(abs(supportvectors2 - supportVectors1)) % This is non zero if Standardize=true

More Answers (0)

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!

Translated by