3.5 Evolutionary Games 99 its outputs are conditional switch rates which describe how frequently players in some class playing a given strategy are considering switching strategies, given the current expected cost vector and subpopulation state. This revision of pure strate- gies is flexible enough to incorporate a wide variety of paradigms, including those based on learning, imitation, adaptation, optimization, etc. The revision of pure strategies describes the procedures that players follow in adapting their behavior to the evolving environment, such as in evolving networks (Internet traffic, flow control, etc.). 3.5.2 Pairwise Interaction Model 3.5.2.1 The Single Population Case Consider a homogeneous large population of players in which the players are ran- domly matched: each player j plays a symmetric bi-matrix game against some randomly selected player j . Every player can select an action from the finite set A. If a player j with action a meets another player j with the action a , player j gets the utility U(a, a ). Denote by x the frequencies of use of actions, i.e., the popula- tion profile. The vector x = (x a ) aA is a probability distribution on A. The number x can be interpreted as the probability with which a given player in the popu-