The new algorithm solves these two problems in “almost linear” time, which means that the algorithm’s runtime is roughly proportional to the amount of time it takes merely to write down the details of ...

We introduce a randomized algorithm for overdetermined linear least-squares regression. Given an arbitrary full-rank m × n matrix A with m ≥ n, any m × 1 vector b, and any positive real number ε, the ...

We give a branch-and-cut algorithm for solving linear programs (LPs) with continuous separable piecewise-linear cost functions (PLFs). Models for PLFs use continuous variables in special-ordered sets ...