hyperSpec-HPC: Fast Sparse Spectral Kernels in R and Rust

The project will develop a high-performance Rust backend for graph-based spatial smoothing of hyperspectral images, delivered through an R package, hyperspec.rust. The package will bridge R’s dgCMatrix format to Rust’s faer sparse matrix structure, establishing a reusable FFI pattern for R–Rust interoperability. On the Rust side, pixel neighborhood graphs will be constructed using petgraph, with configurable connectivity to minimize data transfer. Spatial smoothing will be implemented by solving the sparse linear system (I+αL)x=b using iterative solvers (CG/BiCGSTAB) from faer. The project will include tests and benchmarking against pure-R implementations on large datasets along with documentation and a vignette to demonstrate smoothing. Additional goals include CRAN release of hyperSpec.rust, publishing the FFI bridge pattern as a Rust crate, and a JOSS paper.