blank space
blank space   blank space
blank space
blank space
blank space
blank space
circle of science
menu bar
menu bar
menu bar
menu bar
menu bar
menu bar
menu bar
menu bar
menu bar
blank space
 RESEARCH  CORE FACILITIES  COMPUTATIONAL MODELING
blank space  
blank space blank space
 

Core Facility for Computational Modeling
Marc Joanisse
Core Facility Homepage (http://www.cllrnet.ca/polaris)

 OBJECTIVES, RATIONALE, HYPOTHESES:

Connectionist models use artificial neural networks, typically implemented as a computer simulation, to study the link between brain and behaviour. Connectionist networks encode information as distributed patterns of activation across groups of artificial neurons, and learn to process information through gradual changes in network connections as a result of experience. These networks have provided considerable insights into the neural structures that may underly a variety of cognitive behaviours, including children's acquisition of morphology (Rumelhart & McClelland 1986), syntax (Elman 1993) and reading (Plaut et al. 1996). They have also contributed to our understanding of various cognitive and language deficits, including particularly dyslexia (Harm & Seidenberg 1999) and developmental language impairments (Joanisse 2000).

 METHODOLOGY:

To provide CLLRnet researchers with a low-cost entry point into this emerging area of research, we will establish a centre dedicated to the computational modeling of cognitive development. The centre will give CLLRnet investigators access to computational modeling expertise at the highest international levels.

 SIGNIFICANCE AND DELIVERABLES:

By creating a centralized base of expertise, collaboration and hardware facilities, connectionist ("neural network") simulations of data pertaining to any project in the network can be developed. The project will be headed by Dr. Marc Joanisse, a cognitive neuroscientist with extensive knowledge in both connectionist models and empirical research in language development. It will include a network of high-performance computer workstations that can be accessed through the Internet to support modeling activities at any CLLRnet node.

The modeling research group will collaborate with researchers throughout the network to facilitate the development and application of models to data from CLLRnet projects. Close collaboration will allow modeling projects to evolve in parallel with the empirical projects. This is a significant departure from and major improvement over the way that modeling has typically been undertaken B at arms length from the actual empirical data.

The project will also train researchers on the types of techniques necessary for developing connectionist simulations. This includes training graduate research assistants and postdoctoral fellows to develop simulations of cognitive processes. Finally, we will create a network of high performance computers for developing and running simulations.

For more information on this project, please contact Marc Joanisse (marcj@uwo.ca.)


Last Modified: July 11 2002 08:49:44.

blank space search link blank space | | grahpic arrow