Ego-centric and allo-centric abstraction in self-organized hierarchical neural networks

The computational systems supporting the cognitive capacity of artificial agents are often structured hierarchically, with sensory-motor details placed in the lower levels, and abstracted conceptual items in the upper levels. Such an architecture mimics the structural properties of the animal and human nervous system. To operate efficiently in varying circumstances, artificial agents are necessary to consider both ego-centric (i.e. self-centered) and allo-centric (i.e. other-centered) information, which are further combined to address given tasks. The present work investigates effective assemblies for simultaneously placing ego-centric and allo-centric processes in the cognitive hierarchy, by evolving self-organized neural network controllers. The systematic study of the internal network mechanisms has showed that effective neural assemblies are developed by placing allo-centric information in the upper levels of the cognitive hierarchy, followed by ego-centric abstracted representations in the middle and finally sensory-motor details in the lower level. We present and discuss the obtained results considering how they are related with known assumptions about human brain functionality.

[1]  Jun Tani,et al.  How Hierarchical Control Self-organizes in Artificial Adaptive Systems , 2005, Adapt. Behav..

[2]  Daphna Joel,et al.  Electrolytic lesions of the medial prefrontal cortex in rats disrupt performance on an analog of the Wisconsin Card Sorting Test, but do not disrupt latent inhibition: implications for animal models of schizophrenia , 1997, Behavioural Brain Research.

[3]  Randall D. Beer,et al.  Spatial learning for navigation in dynamic environments , 1996, IEEE Trans. Syst. Man Cybern. Part B.

[4]  Gregory Z. Grudic,et al.  Terrain Segmentation with On-Line Mixtures of Experts for Autonomous Robot Navigation , 2009, MCS.

[5]  Jun Tani,et al.  Emergence of Functional Hierarchy in a Multiple Timescale Neural Network Model: A Humanoid Robot Experiment , 2008, PLoS Comput. Biol..

[6]  Randall D. Beer,et al.  On the Dynamics of Small Continuous-Time Recurrent Neural Networks , 1995, Adapt. Behav..

[7]  Matthew S Stanford,et al.  Latent structure of the Wisconsin Card Sorting Test: a confirmatory factor analytic study. , 2005, Archives of clinical neuropsychology : the official journal of the National Academy of Neuropsychologists.

[8]  Jeffrey L. Krichmar,et al.  Neuromodulation as a robot controller , 2009, IEEE Robotics & Automation Magazine.

[9]  P. Mandik.  PHENOMENAL CONSCIOUSNESS AND THE ALLOCENTRIC-EGOCENTRIC INTERFACE , 2005 .

[10]  Daniel L. Silver,et al.  Inductive transfer with context-sensitive neural networks , 2008, Machine Learning.

[11]  Shai Ben-David,et al.  Exploiting Task Relatedness for Mulitple Task Learning , 2003, COLT.

[12]  Panos E. Trahanias,et al.  Explorations on artificial time perception , 2009, Neural Networks.

[13]  Tom Ziemke,et al.  Neuromodulation of Reactive Sensorimotor Mappings as a Short-Term Memory Mechanism in Delayed Response Tasks , 2002, Adapt. Behav..

[14]  Roy M. Turner,et al.  Context-Sensitive Weights for a Neural Network , 2003, CONTEXT.

[15]  Jun Tani,et al.  Acquiring Rules for Rules: Neuro-Dynamical Systems Account for Meta-Cognition , 2009, Adapt. Behav..

[16]  Panos E. Trahanias,et al.  Self-organized executive control functions , 2010, The 2010 International Joint Conference on Neural Networks (IJCNN).

[17]  Aude Billard,et al.  Neural Model of the Transformation from Allo-centric to Ego- centric Representation of Motions , 2005 .