Cognitive Networks: Towards Self-Aware Networks

9.6. Conclusion

This chapter has discussed distributed learning and reasoning for cognitive networks. We have shown how a cognitive network fits into the multiagent system and cooperative distributed problem-solving contexts. We have described the application of three classes of distributed reasoning methods to a cognitive network and provided justification for their use, as well as potential drawbacks. Each class of distributed reasoning methods has been paired with a distributed learning method, and the advantages and disadvantages of each learning method have been described. Sensor and actuator functions have been presented as an integral part of learning and reasoning in a cognitive network. Finally, we have investigated behavior, computational state and cognitive control as areas encompassing important design decisions.

Cognitive networking is rich with opportunities for novel research. The areas of learning and reasoning are particularly ripe and central to the success of cognitive network implementations. Most distributed learning and reasoning methods described in this chapter assume that communication is reliable and error-free and that information obtained is accurate. In real-world networks, particularly wireless ones, ensuring reliable communication may be difficult. Also, information may be noisy, inaccurate or purposely falsified by malicious nodes. Thus, a major opportunity exists for developing learning and reasoning methods that are robust and that incorporate ...