Output details
11 - Computer Science and Informatics
University of Birmingham
When is 'nearest neighbour' meaningful: A converse theorem and implications
<24>This paper elucidates a subtle aspect of the curse of dimensionality by proving that the previously known sufficient conditions for distance concentration are in fact necessary. Early draft version contributed to the MRC Grant G0701858. Third most downloaded Journal of Complexity paper on Science Direct TOP25 Hottest Articles in 2009/10. Sparked new algorithm development by others at high impact venues, including JMLR. Motivated the development of a new statistical test to detect meaningless distances (A Kaban, Statistics & Computing 22, 2, 375-385, 2012) which a referee found "of great practical use and of high significance for high dimensional data analysis."