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 asndarray::Array2
orkmedoids::arrayadapter::LowerTriangle
- type
N
- number data type such asu32
orf64
- type
L
- number data type such asi64
orf64
for the loss (must be signed) mat
- a pairwise distance matrixmed
- the list of medoidsmaxiter
- 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);