Stochastic learning algorithms for optimal design of wireless networks

We introduce algorithms to optimize wireless networks in the presence of fading. Central to the problem considered is the need to learn the fading's probability distribution while determining optimal operating points. A stochastic subgradient descent algorithm in the dual domain is developed to accomplish this task. Even though the optimization problems considered are not convex, convergence of the proposed algorithms is claimed. Numerical results using adaptive modulation over an interference limited physical layer corroborate theoretical results.