Landau Theory of Meta-learning

Abstract

Computational Intelligence (CI) is a sub-branch of Artificial Intelligence paradigm focusing on the study of adaptive mechanisms to enable or facilitate intelligent behavior in complex and changing environments. Several paradigms of CI [like artificial neural networks, evolutionary computations, swarm intelligence, artificial immune systems, fuzzy systems and many others] are not yet unified in the common theoretical framework. Moreover, most of those paradigms evolved into separate machine learning (ML) techniques, where probabilistic methods are used complementary with CI techniques in order to effectively combine elements of learning, adaptation, evolution and Fuzzy logic to create heuristic algorithms. The current trend is to develop meta-learning techniques, since no single machine learning algorithm is superior to others in all-possible situations. The mean-field theory is reviewed here, as the promising analytical approach that can be used for unifying results of independent ML methods into single prediction, i.e. the meta-learning solution. The Landau approximation moreover describes the adaptive integration of information acquired from semi-infinite ensemble of independent learning agents, where only local interactions are considered. The influence of each individual agent on its neighbors is described within the well-known social impact theory. The final decision outcome for the meta-learning universal CI system is calculated using majority rule in the stationary limit, yet the minority solutions can survive inside the majority population, as the complex intermittent clusters of opposite opinion.