ABSTRACT
This chapter aims to combine the two approaches of various classes of strategy learning schemes and payoff-learning schemes. It discusses combined learning in dynamic games with incomplete information and imperfect payoffs. The chapter considers fundamental issues in the study of dynamic interactive systems and presents the model and notations. It analyzes the convergence to novel game dynamics based on heterogeneous learning. The chapter provides an illustrative example of robust games and learning in wireless networking. In many dynamic interactions, it is convenient to have a learning and adaptive procedure that works with minimal information and less memory as possible. The chapter aims to develop a fully distributed learning procedure to learn the expected payoffs as well as the optimal strategies associated with the learned payoffs. It outlines a mathematical program to find the fastest learning algorithms. The program suggests looking at the learning patterns that have second highest eigenvalues, which are as small as possible.
