Nearest Neighbours

Question 1

What type of method is nearest neighbours?

Answer 1

Non-parametric

Question 2

How does nearest neighbours work?

Answer 2

We store a large amount of training data. When the AI is fed a test sample, it compares it with the training data to find the closest match

Question 3

What data is required for nearest neighbours?

Answer 3

We need to store and keep all data

Question 4

What does “k-nearest” mean?

Answer 4

When finind the nearest matches, we can average k number of close points to get a better average

Question 5

How does nearest neighbours scale with dimensions?

Answer 5

In few dimensions, it works well. In larger dimensions, it is very hard.

Question 6

How does k affect the boundary?

Answer 6

> A low value of k makes a much rougher boundary

> A high value of k makes a much smoother boundary

[Picture 5]

Question 7

What is a KD-tree?

Answer 7

For finding neighbours in a large number of dimensions, we use a KD tree. The tree can be easily navigated to find the point