Skip to contents

Leading eigenvector decomposition (LEiDA) – C++ / LAPACK backend

Source: R/RcppExports.R
get_leida_cpp.Rd

Internal workhorse called by get_leida(). For each phase-locking matrix slice in sync_conn, calls LAPACK dsyev (symmetric eigendecomposition) and returns only the leading eigenvector.

Usage

get_leida_cpp(sync_conn)

Arguments

sync_conn

NumericVector with dim attribute [N, N, t_points].

Value

NumericMatrix [t_points x N]. Leading eigenvectors, one per row.

Details

Differences from the R implementation:

  • dsyev returns eigenvalues in ascending order, so the leading eigenvector is the last column of the output (opposite of R's eigen() which is descending).

  • The LAPACK workspace is allocated once and reused across all timepoints.

  • Sign convention preserved: row sum forced to be non-positive.