Hello,
How did they find the boundaries for qu4? And also, for qu3, i don't get at all, why they take k=5 or 7 instead in the leave one out cross validation? Taking k=1 would lead to a zero error isn't it? Or other k that are still less than 14 and and odd would work in my opinion!
To sketch the boundary, you should draw lines that are always the same distance to two points from different classes. You could approach the problem by first asking yourself "for each area of the space, which two points are the closest?" If those are from different datasets, you should find the middle-line.
If you choose k=1, you get 5/7 of the points wrong. For each cloud of (2+5) points, all except the two outermost points are closest to a point of the opposite class.
Pb18 2018
Hello,
How did they find the boundaries for qu4? And also, for qu3, i don't get at all, why they take k=5 or 7 instead in the leave one out cross validation? Taking k=1 would lead to a zero error isn't it? Or other k that are still less than 14 and and odd would work in my opinion!
To sketch the boundary, you should draw lines that are always the same distance to two points from different classes. You could approach the problem by first asking yourself "for each area of the space, which two points are the closest?" If those are from different datasets, you should find the middle-line.
If you choose k=1, you get 5/7 of the points wrong. For each cloud of (2+5) points, all except the two outermost points are closest to a point of the opposite class.
Add comment