RandNLA

Numerical Linear Algebra drives modern computation, especially since the deep learning revolution. There has been a push to design hardware for the sake of speeding up these computations. On the algorithmic side of things, there has been a small community of researchers that has shown that randomisation can offer massive speedups at the cost of little accuracy, with a recent outlining the creation of RandLAPACK and RandBLAS, randomised counterparts to the popular BLAS and LAPACK libpaperraries. This project is an attempt to bring these developments to the Rust ecosytem. A full technical report can be found here, with the code sitting here