Coordinate descent minimizes an objective by repeatedly optimizing one coordinate, or one block of coordinates, while holding all others fixed. Exact coordinate minimization is especially simple for separable penalties and quadratic objectives.
New to topics? Read the docs here!