Hello,
I'm trying to implement a SVM in python with numpy but I'm having a hard time doing so.
I've been using this link: http://www.tristanfletcher.co.uk/SVM%20Explained.pdf to understand the theory.
However I'm unsure what kind of multiplications they're applying to their matrices. I'm only familiar with the dot product in the context of 2 vectors, but I've read that standard matrix multiplications can be called dot products. So I think their "." operator refers to that. But then I'm unsure what kind of multiplications they're applying to the other matrices.
Page 5 formula 1.13 to calculate H they do sum(sum(y*y*x.x)), how am I supposed to implement that? I've looked at implementations where numpy.outer was used but I'm unsure why...
I've also tried calculating them with 2 for loops, element by element, and then tried finding a formula that'd give the same result but I end up with sum( sum( Y*(Y*X.T).dot( (X.T) ) ) ) which doesn't look anything like what's written in the tutorial.
So in conclusion I'd be very grateful for any help.
[link][5 comments]