A method, framework, and tutorial for efficiently simulating models of decision-making.

Area of Science:

• Cognitive Science
• Computational Neuroscience
• Psychology

Background:

• Evidence accumulation models (EAMs) are key for understanding rapid decision-making.
• Complex EAM variants offer greater explanatory power but are computationally intensive.
• Existing fitting methods are hindered by simulation costs and intractable likelihoods.

Purpose of the Study:

• To present a novel framework for efficiently fitting complex EAMs.
• To introduce an optimized simulation method for EAMs.
• To facilitate broader application of sophisticated decision-making models.

Main Methods:

• Developed a new simulation technique using look-up tables (LUTs) for inverse cumulative density function (iCDF) random number generation (RNG), termed LUT-iCDF.
• Implemented C code for 12 EAM variants utilizing the LUT-iCDF method.
• Validated the accuracy of LUT-iCDF simulations against standard RNG methods in R.

Main Results:

• Identified random number generation as a bottleneck in complex EAM simulations.
• Demonstrated that LUT-iCDF closely approximates standard RNG methods with appropriate LUT sizing.
• Achieved significant speed-ups in simulation efficiency for complex EAMs.

Conclusions:

• The proposed framework and LUT-iCDF method substantially reduce computational costs for complex EAMs.
• This facilitates more accessible and efficient application of advanced decision-making models.
• The provided tutorial and code enable easier implementation of complex EAMs in research.