`mod pca`¶

module pca¶

pca command: streaming PCA of the depth matrix.

Functions

fn rank1_upper_update_sparse(gram: &mut [f64], mean: &mut [f64], depths: &[u16], n: usize)¶: Sparse rank-1 outer product for highly zero depth rows (exact). Exposed for microbenchmarks; production PCA uses the dense inlined loop above.

fn run(params: &PcaParams) -> Result<(), Box<dyn std::error::Error>>¶

fn run_to_arrow(params: &PcaParams) -> Result<PcaArrowResult, Box<dyn std::error::Error>>¶

fn run_to_arrow_with_source<S: MarkerStream>(source: &S, params: &PcaParams) -> Result<PcaArrowResult, Box<dyn std::error::Error>>¶

fn run_with_source<S: MarkerStream>(source: &S, params: &PcaParams) -> Result<(), Box<dyn std::error::Error>>¶

Structs and Unions

struct PcaArrowResult¶

eigenvalues: RecordBatch¶

loadings: RecordBatch¶

n_markers: u64¶

n_individuals: usize¶

n_components: usize¶

total_variance: f64¶

struct PcaParams¶

markers_table_path: String¶

output_dir: String¶

min_depth: u16¶

n_components: Option<usize>¶

backend: crate::compute_backend::PValueBackend¶: Where the Gram accumulation runs.