Trace feature map: a model of episodic associative memory

Area of Science:

• Cognitive Science
• Computational Neuroscience
• Artificial Intelligence

Background:

• Human episodic associative memory involves recalling specific past events.
• Existing computational models often struggle to replicate nuanced memory behaviors like interference and recall biases.
• Topological feature maps offer a structured way to represent data, potentially applicable to memory modeling.

Purpose of the Study:

• To present a novel computational approach for modeling human episodic associative memory.
• To design a memory model with desirable properties such as plausible interference and graceful degradation.
• To leverage topological feature maps for efficient episodic memory representation and retrieval.

Main Methods:

• Utilizing topological feature maps to represent data, creating "trace feature maps".
• Implementing episodic storage where single presentations generate memory traces.
• Developing retrieval mechanisms using partial cues and incorporating overlap for interference.
• Analyzing recall performance based on recency and uniqueness of traces.

Main Results:

• The trace feature map model successfully stores episodic traces from single presentations.
• Partial cues enable effective retrieval of stored episodic information.
• Overlapping traces lead to realistic memory interference patterns.
• Memory performance degrades gradually under overload, with recency and uniqueness influencing recall.

Conclusions:

• The proposed trace feature map approach offers a viable computational model for episodic associative memory.
• The model successfully replicates key human memory characteristics, including interference and recall biases.
• This framework provides a foundation for further research into artificial memory systems and cognitive modeling.