Machine Learning course- Shai Ben-David: Lecture 23

CS 485/685, University of Waterloo. Apr 1, 2015
Last lecture - Support Vector Machines

[albertodelpia]

Thank you Professor for this fantastic course. I binge watched it in one week. Better than Game of Thrones!!

[samw.]

04:11 Dot Product of two vectors
11:27 Suport Vector Machine
38:30 The Kernel Trick
01:10:31 RBF Kernels
01:14:17 K-Nearest Neighbour Learning

[davide]

7:19 I would like to make a comment to the drawing, as a student who has always struggled to remember why that distance is computed like this.

My confusion arises from the fact that the vector w is often drawn to be as long as the distance, which is not the case.

At first it always triggers my question "if ||w|| is 1, how can the distance not be 1?".


In my opinon it would help if the w was drawn of arbitrary length, and the segment d was added beside it.
Then it would be clearer that <w, x> = ||w|| ||x|| cos theta, and d = ||x|| cos theta.

Thank you for this course. Extremely useful and clear.