pub fn fastpam1<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 FastPAM1 algorithm, which yields the same results as the original PAM.

This is faster than PAM, but slower than FasterPAM, and mostly of interest for academic reasons. Quality-wise, FasterPAM is not worse on average, but much faster.

This is the improved version from the journal version of the paper, which costs O(n²) per iteration to find the best swap.

  • 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

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

Example

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::fastpam1(&data, &mut meds, 100);
println!("Loss is: {}", loss);