Quote:
Originally Posted by Michael Reach
I wonder if I'm missing something  I found this lecture to be by far the hardest to understand till now

Hi,
SVM does not build on validation. It is a new classification algoithm for linearly separable data with one added twist: maximizing the margin. May I suggest that you look at the first 6 minutes of this segment again:
If you are sold on the idea of maximizing the margin as something beneficial, what the rest of the lecture does is go through the mathematical and algorithmic machinery of maximizing the margin. After that, we do the usual; apply nonlinear transforms to generalize a linear method. I will be happy to discuss this further.