Greedy k-center algorithm for diverse sample selection.
Iteratively selects points that are farthest from all previously
selected points.
- Parameters:
features (Tensor) – Feature representations (N, D)
k (int) – Number of centers to select (clamped to N if larger)
initial_centers (Optional[Tensor]) – Optional initial centers
- Return type:
Tensor
- Returns:
Indices of selected centers
