1
$\begingroup$

I started reading up on SVM and very little is defined of what are support values. I reckon it's they are denoted as $\alpha$ in most formulations.

$\endgroup$
3
  • 3
    $\begingroup$ Maybe you should provide some context, for example, the formula you're referring to and the links to the article(s) you're reading (although, in this case, it should not be necessary, but still). $\endgroup$ – nbro Jan 5 at 19:39
  • 1
    $\begingroup$ Hello! Welcome to AI.SE. Support vectors are basically an alternative view of SVMs. We view the data points as vectors and weighing them by $\alpha_i$'s will give us the classifier (this is called the representer theorem). Using this we can develop intersting theories about SVM. Also the $\alpha$ is the same as the one that comes up in the Lagarngian, this set of $\alpha_i$'s will act as the solution of your dual problem (a term you have to solve in a convex optimization problem).I cannot elaborate more as this requires some knowledge about linear algebra and optimization.But this is the idea. $\endgroup$ – DuttaA Jan 6 at 19:36
  • 1
    $\begingroup$ Don't think too much of it if you are just trying to implement SVMs but if you are researching about SVM you have to take graduate level Linear Algebra and Convex optimization course to undersatnd this. $\endgroup$ – DuttaA Jan 6 at 19:40
1
$\begingroup$

In the least-squares SVM (LS-SVM) the non-zero Lagrange multipliers ($\alpha$) are the support values. The corresponding data points are the support vectors. Johan Suykens explains this in Least Squares Support Vector Machines.

$\endgroup$
1
  • $\begingroup$ Thank you. The paper is really comprehensive $\endgroup$ – axelmukwena Jan 9 at 12:49

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.