[source] # Canberra A weighted version of the [Manhattan](manhattan.md) 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. $$ Canberra(\mathbf {a} ,\mathbf {b} )=\sum _{i=1}^{n}{\frac {|a_{i}-b_{i}|}{|a_{i}|+|b_{i}|}} $$ **Data Type Compatibility:** Continuous ## Parameters This kernel does not have any parameters. ## Example ```php use Rubix\ML\Kernels\Distance\Canberra; $kernel = new Canberra(); ``` ## References [^1]: G. N. Lance et al. (1967). Mixed-data classificatory programs I. Agglomerative Systems.