reconstruct_jacobian

scirs2_optimize::sparse_numdiff::compression

Function reconstruct_jacobian 

Source 
pub fn reconstruct_jacobian(
    gradients: &ArrayView2<'_, f64>,
    b: &ArrayView2<'_, f64>,
    _c: &ArrayView2<'_, f64>,
    sparsity: &CsrArray<f64>,
) -> Result<CsrArray<f64>, OptimizeError>
Expand description

Reconstructs a sparse Jacobian from compressed gradient evaluations

Takes the compressed gradient evaluations and reconstructs the full sparse Jacobian using the compression matrices. Each compressed gradient contains information about multiple columns that were grouped together during compression.

§Arguments

  • gradients - Matrix of compressed gradient evaluations (m x num_colors)
  • b - Column compression matrix (n x num_colors)
  • c - Row compression matrix (m x m), typically identity for Jacobian
  • sparsity - Original sparsity pattern for reconstruction

§Returns

  • Reconstructed sparse Jacobian