Dear Professor Gerstner,
I'm glad you're feeling better!
I posted in the discussion on courseware yesterday. I don't know if anyone is checking the discussion on courseware. I mean, I'll stick to moodle now that we're back on moodle :)
Anyway, my post was just a small remark about how the Closest Pattern (aka nearest neighbor) algorithm will generally fail for high-dimensional inputs such as vectorized images. This "Curse of Dimensionality" arises because points in high-dimensional space don't have any neighbors close-by unless the number of points grows exponentially with the number of dimensions. So, in my view, it might not be the best example to use at that point in the lecture, or maybe just the type of input (vectorized image) could be replaced by a lower-dimensional type of input.
Best regards,
Felix Grimberg
P.S.: I fall into category (ii): The lectures were very clear and so I did not have any questions about the pen&paper exercises!
