Cognitive ecology and social learning inspired machine learning: with particular reference to the evolving of resilient Airborne Networks (AN)

Abstract

This paper tries to bring together three seemingly only remotely related fields: animal cognition in biology, machine learning in computer science, and the planning and deployment of resilient Airborne Networks. The underlying motivation is that the latest advances in animal behavior ecology such as social learning, innovation, and cognitive ecology may offer some meaningful insights for computational intelligence such as machine learning. Motivated by this expectation, I first review some of the latest advances in the field of animal cognition, with focusing on social learning, teaching, innovation and cognitive ecology. The justification for this focus is not only because they are interesting and are among the most actively studied topics in behavior biology, but also because the existing machine learning research, which from time to time takes cues from cognitive science, seems to only incorporate the traditional learning theory such as associative learning and reinforcement learning. After a briefly review of the major advances in animal learning and cognitive ecology, I look into the possibility to incorporate the principles and mechanisms from social learning and cognitive ecology into a typical machine learning architecture. By examining the Gadanho's (2001, 2003) ALEC (Asynchronous Learning by Emotion and Cognition) architecture, I propose to add a high layer to the ALEC architecture, and the resulting CEML (Cognitive Ecology and social learning inspired Machine Learning) offers a framework that uses a population of agents and can readily consider social learning, teaching, and innovation as well as the influences of environment in a comprehensive manner. Finally, I consider the problem of planning and deployment of Airborne Networks (AN) and suggests that the new CEML should be ideal for tackling the AN problem.