In this thesis manuscript, the main objective is to study the wireless networks where the node terminals are equipped with multiple antennas. Rising topics such as self-optimizing networks, green communications and distributed algorithms have been approached mainly from a theoretical perspective. To this aim, we have used a diversified spectrum of tools from Game Theory, Information Theory, Random Matrix Theory and Learning Theory in Games. We start our analysis with the study of the power allocation problem in distributed networks. The transmitters are assumed to be autonomous and capable of allocating their powers to optimize their Shannon achievable rates. A non-cooperative game theoretical framework is used to investigate the solution to this problem. Distributed algorithms which converge towards the optimal solution, i.e. the Nash equilibrium, have been proposed. Two different approaches have been applied: iterative algorithms based on the best-response correspondence and reinforcement learning algorithms. Another major issue is related to the energy-efficiency aspect of the communication. In order to achieve high transmission rates, the power consumption is also high. In networks where the power consumption is the bottleneck, the Shannon achievable rate is no longer suitable performance metric. This is why we have also addressed the problem of optimizing an energy-efficiency function.
