Mean-Field Theory of Meta-Learning

We discuss the mean-field theory for a cellular automata model of meta-learning. The system is build from an ensemble of interacting, learning agents, that acquire and process incoming information using various types, or versions of machine learning algorithms. The abstract learning space, where all agents are located, is constructed here using a fully connected model that couples all agents with random strength values. The cellular automata type of evolution simulates here the higher level integration of information acquired from the ensemble of independent learning agents. The final classification of incoming input data is defined as the stationary state of the meta-learning system using simple majority rule, yet the minority clusters that share opposite classification outcome can be observed in the system. Therefore, the probability of selecting proper class for a given input data, can be estimated even without the actual knowledge of its affiliation. Therefore, fuzzy logic can be easily introduced into the system, even if learning agents are build from simple, binary classification agents.