Learning and Evolution by Minimization of Mutual Information

Based on negative correlation learning [1] and evolutionary learning, evolutionary ensembles with negative correlation learning (EENCL) was proposed for learning and designing of neural network ensembles [2]. The idea of EENCL is to regard the population of neural networks as an ensemble, and the evolutionary process as the design of neural network ensembles. EENCL used a fitness sharing based on the covering set. Such fitness sharing did not make accurate measurement on the similarity in the population. In this paper, a fitness sharing scheme based on mutual information is introduced in EENCL to evolve a diverse and cooperative population. The effectiveness of such evolutionary learning approach was tested on two real-world problems. This paper has also analyzed negative correlation learning in terms of mutual information on a regression task in the different noise conditions.