Project : Cortical Software Re-Use :: Computational Cognitive Modeling Research Group, Ireland

Funded by:

Basic Research programme of Enterprise Ireland

Duration:

Oct 1999 - June 2002

Project Leader:

Dr. Ronan Reilly

Collaborators:

Dr. M-Tahar Kechadi (Dept. of Computer Science, University College Dublin);
Dr. Ted Cox, Dr. Dana Nicolau (Dept. of Maths Physics, University College Dublin);
Dr. Teresa Burke (Dept. of Psychology, University College Dublin)

Postgraduate researchers:

Eithne Shalloo (MSc student)
Ioana Marian (MSc student)

Description:

The Cortical Software Re-Use (CSRU) project is an interdisciplinary project which involved the Departments of Computer Science, Psychology, and Mathematical Physics at UCD. The motivation behind this project was to develop the mathematical basis for, and explore the psychological implications of, a theory developed by the principal investigator (Reilly, 1997). The theory in question is that of aims to account for a range of neural computational and cognitive development phenomena. The central concept of the theory, that of "software re-use," is borrowed from the field of software egineering. Put simply, it states that dynamical neural processes from the sensory-motor areas of the brain provide the computational building blocks for higher level functions up to and including those involved in cognition and language. The specific goals of the project were:

to provide a mathematical foundation for the theory of CSRU
to demonstrate the feasibility of CSRU by constructing a large scale computational model of neuropsychological phenomena, and
to develop an integrated set of mathematical and computational tools for modelling dynamical neural systems exploiting high performance computing techniques such as parallelism

Principles of CSRU: The cortical "algorithms" for language processing and cognition are derived from, and built upon, those from the sensory-motor domain. According to this view, cortical circuits that are involved in the planning of motor movements, say, can be exploited during reasoning, and not necessarily when reasoning about movement. The functioning of these algorithms is mediated by reciprocal projections between sensory, motor, association, and prefrontal areas. The direction of influence is from the sensory and motor regions to the higher cortical regions, because the circuitry in the sensory-motor areas consolidates and matures earlier, and that of the prefrontal area remains plastic the longest.

The style of computation is a form of dynamical constraint satisfaction, where oscillating patterns of neuronal firing from connected regions mutually influence one another through a process of resonance and harmonisation. In computational terms one can think of two dynamical patterns resonating as being equivalent to the application of a function designed for one domain (e.g., sensory-motor) to a new domain (e.g., cognitive). The nature and quality of the repertoire of sensory-motor algorithms determines the nature and quality of the cognitive and linguistic processes built from them. Consequently, individuals with different sensory motor experiences will have, for example, different cognitive styles.

Outcomes:

The Cortical Software Re-Use project has been completed successfully with most of the project aims met, and some exceeded. A mathematical foundation for CSRU has been developed, and work on this area is continuing beyond the completion of the project. A large scale, neurologically plausible model of visuo-motor imitation has been successfully implemented. This model was developed as an exemplar of the CSRU process. In support of this model's development a high-performance simulation environment was sucessfully developed. Work related to all three of these major goals has led to a number of conference presentations, book chapters, and journal submissions. In addition, the project has given rise to one psychology MA thesis.

Relevant Literature:

Reilly, R.G. (1995). Sandy ideas and coloured days: The computational implications of embodiment. Artificial Intelligence Review, 9, 305-322. (pdf)

Reilly, R.G. (2000). Evolution of symbolization: Signposts to a bridge between connectionist and symbolic systems. In S. Wermter & R. Sun (Eds.), Hybrid neural systems. Berlin, Germany: Springer. (pdf)

Reilly, R.G. (2001). Collaborative cell assemblies: Towards a cortical computational building block. In Proceedings of EmerNet: Third International Workshop on Current Computational Architectures Integrating Neural Networks and Neuroscience. University of Durham, August. (pdf)

Marian, I. and Reilly, R.G. (2001) SpikeSNNS - a simulator for spike processing networks. Proceedings of 12th DAAAM International Symposium on Intelligent Manufacturing & Automation: Focus on Precision Engineering, B. Katalinic (Ed.). DAAAM International, Jena, Germany, October 2001.

Marian, I. and Reilly, R.G. (2001) Self-organization of neurons coding directional selectivity in motor cortex. Proceedings of 12th Irish Conference on Artificial Intelligence and Cognitive Science (AICS), D. O'Donoghue (Ed.). NUI Maynooth, Ireland, September 2001, pp. 253-264.

Marian, I., Reilly, R.G. and Mackey, D. (2002) Efficient event-driven simulation for spiking neural networks. Proceedings of 3rd WSES International Conference on: Neural Networks and Applications. Interlaken, Switzerland, February 2002. (pdf)

Reilly, R.G., Marian, I. (2002). Cortical Software Re-Use: A Computational Principle for Cognitive Development in Robots. Proceedings of ICDL 2002. MIT Press. (pdf)

Reilly, R.G. (In Press). The relationship between object manipulation and language development in Broca's area: A connectionist simulation of Greenfield's hypothesis. Behavioral and Brain Sciences. (pdf)