Abstract:
We present a model for the self-organized formation of hippocampal place cells and limbic head direction cells based on unsupervised learning on natural visual stimuli. The model is based on a hierarchy of Slow Feature Analysis (SFA) modules, which were recently shown to be a good model for complex cells in the early visual system. The system extracts a distributed representation of position and orientation, which is transcoded into a localized place field or head direction representation, respectively, by sparse coding (ICA). We introduce a mathematical framework for determining the solutions of SFA, which accurately predicts the distributed representation of computer simulations. The model's output characteristic of orientation-independent or orientation-dependent place cell-type or position-independent head direction cell-type solely depends on the animal's movement pattern, which explains experimental findings about place cell head direction selectivity depending on an animal's behavioral task.