I am newbie reading Support Vector Machines Explained. Translation to dual Quadratic program (p.3 - p.4) in SVM explaining article I mentioned confused me because I understand basic Lagrange multiplier approach but do not understand duality.
My question is:
Could
$\frac{1}{2} {\parallel\omega\parallel}^{2} - \sum_{i=1}^{L}{\alpha}_{i}[{y}_{i}({x}_{i}\omega - b)-1]$
be solved directly without being translated to dual Lagrange task?