1
$\begingroup$

I am working on bag of words technique for image recognition project and after I run k means I get a distance matrix and I have no idea how to apply that to svm to recognize images ?

$\endgroup$

1 Answer 1

1
$\begingroup$

The common Kernerl SVM requires a (valid) Gram matrix. You can turn your distance matrix to a Kernel matrix and feed it into the SVM to see what it spits.

In general, Kernel matrix are "diagonal heavy"-ey (I'm pretty sure I read that term in this very site actually), in the sense that they do not measure distance, but a sense of similarity (so big distance should imply a small value in the Kernel matrix, while small distances convey large values in the Kernel matrix).


This question provides some references which might guide you in the endeavor: Turn a distance measure into a kernel function

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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