Neural field model for perceptual learning

Perceptual learning happens at the perceptual level. We combine holism and reductionism to research perception learning. Based on information geometry the paper presents a neural field model which is used to understand the transformation mechanism, dynamical behavior, capability and limitation of neural network models, by the study of globally topological and geometrical structure on parameter spaces of neural networks. We discuss neural field representation, fractal learning principle, topology approximation correction learning and the dualistic correction learning algorithm. Finally the paper gives conclusions and points out future research topics.