Echo State Networks for Hierarchical Cognitive Control
Remco Tukker, Anne van Rossum, Pim Haselager, Stefan Leo Frank · 2012
In this thesis, we will present an Echo State Network (ESN) to investigate hierarchical cognitive control, one of the functions of Prefrontal Cortex (PFC). This ESN is designed with the intention to implement it as a robot controller, making it useful for biologically inspired robot control and for embodied and embedded PFC research. We will apply the ESN to a n-back task and a Wisconsin Card Sorting task to confirm the hypothesis that topological mapping of temporal and policy abstraction over the PFC can be explained by the effects of two requirements: a better preservation of information when information is processed in different areas, versus a better integration of information when information is processed in a single area.