Andreas Knoblauch, Rüdiger Kupper, Marc-Oliver Gewaltig, Ursula Körner, Edgar Körner, "Design and simulation of a cortical control architecture for object recognition and representational learning", Proceedings of the 3rd HRI International Workshop, pp. 6–13, 2005.

In this work we present simulation results substantiating a previously proposed model of computation in neocortical architecture (Koerner et al., Neural Networks 12:989-1005, 1999). For each stage of the cortical hierarchy we hypothesize three interacting columnar systems for (A) fast forward recognition, (B) refined recognition by feedback, and (C) behavior/prediction related processes, roughly corresponding to middle (IV), superficial (II/III), and deep (V/VI) cortical layers. In a first example we implement a simple system for word recognition. Focusing on the dynamics, we explore the interaction of the A and B systems in recognizing words, quenching out expected signals, and representing new words based on previously learned representations. In a second large-scale implementation of the visual and saccadic system we additionally demonstrate the learning of new object representations and the generation of predictions within the C system based on saccadic sequences.