Parallel distributed processing and neural networks: origins, methodology and cognitive functions

Area of Science:

• Neuroscience
• Computational Neuroscience
• Cognitive Science

Background:

• Parallel Distributed Processing (PDP) is a computational methodology rooted in Associationism.
• Supercomputers are increasingly used by neuroscientists to model brain behavior relationships.

Purpose of the Study:

• To provide empirical information regarding neurobiological systems using PDP.
• To explore the advantages of parallel over serial processing in supercomputer architecture for brain modeling.

Main Methods:

• Overview of supercomputer architecture.
• Analysis of histological data for evidence of parallel distributed processing in the human brain.
• Computer simulations of brain functions.
• Review of in vivo brain imaging studies on attention and language tasks.

Main Results:

• Supercomputer architecture highlights the benefits of parallel processing over serial processing.
• Histological data and computer simulations support the parallel distributed nature of some human brain aspects.
• While sensory networks may follow sequential pathways, attention and language tasks involve multiple brain areas.
• A shift in interest towards associative models over Artificial Intelligence symbolic models for neural networks.

Conclusions:

• Parallel Distributed Processing models, supported by supercomputing and empirical data, are valuable for understanding neurobiological systems.
• The human brain exhibits parallel distributed processing, particularly in complex cognitive functions like attention and language.
• Associative models are gaining prominence over symbolic AI models in reflecting neural network functions.