File Entry: A Neocortex Inspired Hierarchical Spatio-Temporal Pattern Recognition System

Undergraduate Thesis presented in partial fulfillment of the requirements for the degree of Bachelor of Computer Science.

Abstract. Recently, inspired by our neocortex, an architecture for spatio-temporal pattern recognition has been proposed, the MPF (Memory-Prediction Framework). There already exists some attempts to implement it, like the HTM (Hierarchical Temporal Memory) and the HQSOM (Hierarchical Quilted Self-Organizing Map). The HQSOM lacks some parts of the MPF, which in turn makes it less powerful than it could be. This work aims to provide a step towards a complete implementation of the MPF by using some variants of a self-organizing map (SOM). In order to reach this goal, a new variant of the SOM is proposed: The LoopSOM. The LoopSOM extends the SOM-RSOM pair of the HQSOM with predictive and fallback abilities. The LoopSOM has been integrated to the HQSOM, making it closer to the MPF. Experiments were done in order to verify the results obtained with this modification.