Affine scaling

A variant of Karmarkar's algorithm for the linear program in standard form: Min{cx: x >= 0, Ax = b}, where A has full row rank. Here are the basic steps (of the primal method). Given x > 0, let X = diag(x_j). Estimate dual: y = [AX²A']–1 A[X²]c' and d = c - y'A. Move: x' = x - aX²d'/Max{d_j x_j}, wh...
Found on
No exact match found