Key parameters

Main hyperparameters that influence search and clustering.

Parameter Effect Reference
base_2d_rays Controls angular resolution in 2D (default 32). 3D scales to ~34; higher dimensions use subspaces. ModalBoundaryClustering
direction Selects scan direction to locate the inflection point: center_out or outside_in. ModalBoundaryClustering
scan_radius_factor, scan_steps Set size and resolution of the radial scan around seed points. ModalBoundaryClustering
optim_method Choose between gradient_ascent (default) or trust_region_newton for local maximisation. ShuShu
grad_* Hyperparameters for gradient ascent (rate, iterations, tolerances); used when optim_method="gradient_ascent". ShuShu
max_subspaces Maximum number of subspaces considered when dimensionality > 3. SubspaceScout
density_alpha, density_k Optional density penalty using HNSW k‑NN to keep centres inside the data cloud; disable with density_alpha=0. ModalBoundaryClustering

Defaults favour speed; see Highlights for performance tips and Limitations for known constraints.