Function kmedoids::fasterpam

source ·
pub fn fasterpam<M, N, L>(
    mat: &M,
    med: &mut Vec<usize>,
    maxiter: usize
) -> (L, Vec<usize>, usize, usize)where
    N: Zero + PartialOrd + Copy,
    L: AddAssign + Signed + Zero + PartialOrd + Copy + From<N>,
    M: ArrayAdapter<N>,
Expand description

Run the FasterPAM algorithm.

If used multiple times, it is better to additionally shuffle the input data, to increase randomness of the solutions found and hence increase the chance of finding a better solution.

  • type M - matrix data type such as ndarray::Array2 or kmedoids::arrayadapter::LowerTriangle
  • type N - number data type such as u32 or f64
  • type L - number data type such as i64 or f64 for the loss (must be signed)
  • mat - a pairwise distance matrix
  • med - the list of medoids
  • maxiter - the maximum number of iterations allowed

returns a tuple containing:

  • the final loss
  • the final cluster assignment
  • the number of iterations needed
  • the number of swaps performed


  • panics when the dissimilarity matrix is not square
  • panics when k is 0 or larger than N


Given a dissimilarity matrix of size 4 x 4, use:

let data = ndarray::arr2(&[[0,1,2,3],[1,0,4,5],[2,4,0,6],[3,5,6,0]]);
let mut meds = kmedoids::random_initialization(4, 2, &mut rand::thread_rng());
let (loss, assi, n_iter, n_swap): (f64, _, _, _) = kmedoids::fasterpam(&data, &mut meds, 100);
println!("Loss is: {}", loss);