Hierarchical Concept Formation in Associative Memory Composed of Neuro-window Elements

Area of Science:

• Computational neuroscience
• Artificial intelligence
• Cognitive modeling

Background:

• Auto-associative memory models are crucial for understanding information storage and retrieval.
• Hierarchical structures are fundamental to human concept formation and learning.
• Neuro-window elements offer a novel approach to neural network design.

Purpose of the Study:

• To propose a model for hierarchical concept formation in artificial neural networks.
• To investigate the emergence of concept patterns from correlated memory data.
• To demonstrate selective retrieval of memory and concept patterns.

Main Methods:

• Development of a computational model using neuro-window elements.
• Implementation of a Hebbian learning rule for pattern association.
• Hierarchical organization of correlated memory patterns.
• Parameter adjustment for selective pattern retrieval.

Main Results:

• Concept patterns representing similar memory patterns were successfully formed.
• Hebbian learning facilitated concept pattern emergence in a mutually connected network.
• Selective retrieval of both memory and concept patterns was achieved by adjusting neuro-window parameters.

Conclusions:

• The proposed model effectively demonstrates hierarchical concept formation in auto-associative memory.
• Neuro-window elements provide a viable mechanism for learning and retrieving complex data structures.
• This work contributes to understanding neural network capabilities in cognitive processes.