Application of gradient algorithm for optimizing power allocation in DSL systems

The optimization of power allocation in DSL systems is a well studied non-convex problem. In recent years many algorithms have been proposed to solve this problem, but due to its non-convexity, researchers had to rely on heuristic, or exhaustive research methods. In our approach we make use of the similarities of the channel gain between 2 adjacent tones to overcome the non-convexity limitation of gradient methods. This allows the proposition of an efficient and a fast routine that manages to speed up the convergence of existent state of the art algorithms.