Compute UMAP Embedding from a kNN Graph
Usage
umap_from_knn(adj, n_neighbors = 5, seed = 1024)
Arguments
- adj
A sparse adjacency matrix (class `dgCMatrix`) representing a kNN graph.
- n_neighbors
Integer. Minimum number of neighbors required per node (default: 5).
- seed
Integer. Random seed for UMAP initialization (default: 1024).
Value
A data frame with columns `umap_1` and `umap_2` and rownames matching input adjacency matrix.
Details
Runs UMAP on a filtered kNN adjacency matrix, ensuring that all nodes have at least `n_neighbors`.
Assumes the embedding is computed via a Python-based backend (e.g., `umap_from_knn_py()`).
This function filters the graph to remove low-degree nodes before computing UMAP.