1: \begin{abstract}

2: Adaptive networks consist of a collection of agents with adaptation

3: and learning abilities. The agents interact with each other on a

4: local level and diffuse information across the network through their

5: collaborations. In this work, we consider two types of agents:

6: informed agents and uninformed agents. The former receive new data

7: regularly and perform consultation and in-network tasks, while the

8: latter do not collect data and only participate in the consultation

9: tasks. We examine the performance of adaptive networks as a function

10: of the proportion of informed agents and their distribution in

11: space. The results reveal some interesting and surprising trade-offs

12: between convergence rate and mean-square performance. In particular,

13: among other results, it is shown that the performance of adaptive

14: networks does not necessarily improve with a larger proportion of

15: informed agents. Instead, it is established that the larger the

16: proportion of informed agents is, the faster the convergence rate of

17: the network becomes albeit at the expense of some deterioration in

18: mean-square performance. The results further establish that

19: uninformed agents play an important role in determining the

20: steady-state performance of the network, and that it is preferable

21: to keep some of the highly connected agents uninformed. The

22: arguments reveal an important interplay among three factors: the

23: number and distribution of informed agents in the network, the

24: convergence rate of the learning process, and the estimation

25: accuracy in steady-state. Expressions that quantify these relations

26: are derived, and simulations are included to support the theoretical

27: findings. We further apply the results to two models that are widely

28: used to represent behavior over complex networks, namely, the

29: Erdos-Renyi and scale-free models.

30: \end{abstract}