Skip to contents

Generate a CAGRA-style GPU-accelerated kNN graph

rs_cagra_gpu_knn.Rd

Builds a kNN graph from an embedding matrix using the CAGRA algorithm on the wgpu backend. Supports two retrieval modes: direct extraction from the NNDescent graph, or beam search over the pruned CAGRA graph. The former tends to have worse Recall.

Usage

rs_cagra_gpu_knn(embd, cagra_params, extract_knn, seed, verbose)

Arguments

embd

Numeric matrix of embeddings, cells x features.

cagra_params

A named list with the parameters, see params_sc_cagra()

extract_knn

Logical. If TRUE, extracts the kNN graph directly from the NNDescent result (faster, slightly lower precision). If FALSE, runs beam search over the pruned CAGRA graph (slower, higher precision).

seed

Integer. Random seed for reproducibility.

verbose

Logical. Whether to print progress messages.

Value

A named list with:

  • indices - Integer matrix of shape cells x k_query with 0-based neighbour indices.

  • dist - Numeric matrix of shape cells x k_query with distances to the neighbours.

  • dist_metric - Character. The distance metric used.