A weighted version of the Manhattan distance, Canberra examines the sum of a series of fractional differences between two samples. Canberra can be very sensitive when both coordinates are near zero.
This kernel does not have any parameters.
use Rubix\ML\Kernels\Distance\Canberra; $kernel = new Canberra();
- G. N. Lance et al. (1967). Mixed-data classificatory programs I. Agglomerative Systems.