Towards Cognitive Autonomous Networks

5Classic Artificial Intelligence: Tools for Autonomous Reasoning

Stephen Mwanje¹, Marton Kajo², Benedek Schultz³, Kimmo Hatonen⁴, and Ilaria Malanchini⁵

¹Nokia Bell Labs, Munich, Germany

²Technical University of Munich, Munich, Germany

³Nokia Bell Labs, Budapest, Hungary

⁴Nokia Bell Labs, Espoo, Finland

⁵Nokia Bell Labs, Stuttgart, Germany

Network management is an information processing task, which, owing to the complexity of the networks, has traditionally been undertaken by human experts. However, as demonstrated by 3G and 4G Self‐Organizing Networks (SONs), a proper application of cognitive techniques can do a lot in automating network management. Cognition, as defined in Chapter 1, refers to the capability to perceive a signal into a data element and subsequently reasoning over the data element to select an action. As such, cognitive techniques refer to the broad grouping of Artificial Intelligence (AI) related methods and technologies developed for information processing.

Autonomous reasoning requires that the cognitive entity perceives relationships amongst data elements and makes inferences about these elements and their relations to subsequently select the appropriate action. Classic AI techniques were developed with inference engines intended to map human reason and draw humanly relatable decisions from data. Consequently, these classic AI techniques can be very useful in designing automated reasoning engines, also applicable to networks and network management.

This ...