Ken Clarkson

My work has mainly been on geometric algorithms, and in particular on algorithms that have provable properties, but are relatively simple. Randomization is quite useful for this, whether via the Vapnik-Chervonenkis dimension, or using the general framework introduced here. See, for example, a survey on randomized geometric algorithms.

My papers and talks

I've coded up a data structure for nearest neighbor searching in metric spaces, which supports nearest neighbor queries, fixed radius queries, and k'th-nearest-neighbor queries. I've tested it on Euclidean data, strings under Hamming and edit distances, and some bitvector datasets. It gives a pretty good speedup for low-dimensional data, and sometimes high-dimensional data, but isn't going to be as fast as more specialized methods.

Sadly, I haven't managed to:

• Be a former agent of the spider-aliens
• Direct Law and Economics Centers
• Play the banjo in Jimmy Chickenpants
• Be a bashful crooner whose career succeeds using the contributions of his girlfriend, who had received a million dollars from a mysterious reclusive billionaire, and who becomes afraid she has lost him, and to whom he at last returns
• Be an alter ego of Superman,  or of an imaginary Superman

Introducing the Collectibles!
Meet the Collectibles!