pub fn fastermsc<M, N, L>(
mat: &M,
med: &mut Vec<usize>,
maxiter: usize
) -> (L, Vec<usize>, usize, usize)where
N: Zero + PartialOrd + Copy,
L: Float + Signed + AddAssign + From<N> + From<u32>,
M: ArrayAdapter<N>,
Expand description
Run the FasterMSC 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 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::fastermsc(&data, &mut meds, 100);
println!("Loss is: {}", loss);