Analytical Expressions for the REM Model of Recognition Memory

Area of Science:

• Cognitive Psychology
• Computational Neuroscience
• Mathematical Modeling

Background:

• Simulation-based models in cognitive psychology, particularly the REM model of recognition memory, demand substantial computational resources.
• Evaluating model predictions often involves time-intensive simulations, hindering rapid analysis and exploration.

Purpose of the Study:

• To derive an analytic expression for the REM model of recognition memory.
• To significantly reduce the computational burden associated with simulating the REM model.
• To verify the accuracy of the derived analytic form and investigate parameter identifiability.

Main Methods:

• Utilized a Fourier transform technique to derive an analytic expression for the REM model.
• Compared the predictions of the derived analytic form with established findings from Shiffrin and Steyvers (1997).
• Analyzed the dependence of REM predictions on the vector length parameter.

Main Results:

• Successfully derived an analytic form expression for the REM model, drastically cutting simulation time.
• Demonstrated high accuracy of the derived model by achieving close correspondence with previously reported predictions.
• Identified that REM predictions are contingent on the vector length parameter and that parameter identifiability requires fixing one parameter.

Conclusions:

• The analytic derivation offers a computationally efficient alternative for REM model simulations.
• The derived model maintains the predictive accuracy of the original REM model.
• Understanding parameter dependencies is crucial for robust model application and interpretation.