Lower bound on dimension for nearest neighbor classifier to fail at k=1 and pass at k=3

72 Views Asked by At

What is the minimum dimensionality of a dataset of a finite number of points where 1-NN has an accuracy of 0% but 3-NN has an accuracy of 100%.

This is certainly possible in 3 dimensions and my intuition is that an infinite number of points makes this possible in 2 dimensions, hence why I have tagged this in fractals, but not with finite points.

As an extension, what is the lower bound for the dimensionality of a dataset where n-NN has an accuracy of 0% but m-NN has an accuracy of 100%.