select_orthogonal_anchors
select_orthogonal_anchors(
embeddings,
k,
seed_indices=None,
candidate_indices=None,
use_bridges=True,
global_num_bits=12,
seed=42,
)Select k maximally spread anchors using greedy farthest-point sampling.
Achieves ~87% of full-bridge recall with ~22% of anchors by eliminating redundancy in anchor placement.
Parameters
| Name | Type | Description | Default |
|---|---|---|---|
| embeddings | np.ndarray | Normalized embeddings (n_points, dim) | required |
| k | int | Number of anchors to select | required |
| seed_indices | Optional[np.ndarray] | Initial seed points (default: super connectors) | None |
| candidate_indices | Optional[np.ndarray] | Pool to select from (default: bridges or all points) | None |
| use_bridges | bool | If True and candidate_indices is None, use bridge points | True |
| global_num_bits | int | LSH bits for bridge detection | 12 |
| seed | int | Random seed | 42 |
Returns
| Name | Type | Description |
|---|---|---|
| OrthogonalAnchorResult | OrthogonalAnchorResult with selected indices |