Related Experiment Videos
Parameterization of connectionist models
Rafal Bogacz1, Jonathan D Cohen
1Princeton University, Princeton, New Jersey, USA. r.bogacz@bristo.ac.uk
Behavior Research Methods, Instruments, & Computers : a Journal of the Psychonomic Society, Inc
|January 12, 2005
Summary
This study introduces a novel method for optimizing connectionist models by minimizing a cost function, ensuring model outputs closely match empirical data for better cognitive modeling. The approach aids in automatic or manual parameter searches, enhancing model accuracy.
Area of Science:
- Cognitive Science
- Computational Neuroscience
- Machine Learning
Background:
- Connectionist models are widely used to simulate human cognition.
- Estimating model parameters to accurately reflect empirical data is crucial for model validation.
- Existing methods may lack efficiency or automation in parameter estimation.
Purpose of the Study:
- To present a novel method for estimating parameters of connectionist models.
- To ensure model outputs closely align with empirical data.
- To provide a tool that aids in both automatic and manual parameter optimization.
Main Methods:
- Minimization of a cost function comparing model output statistics to subject performance statistics.
- Utilizing an optimization algorithm to identify parameters that minimize the cost function.
- The cost function also assesses the statistical significance of differences between model and data.
Main Results:
- The proposed method effectively estimates connectionist model parameters.
- The cost function quantifies the discrepancy between model and empirical data statistics.
- The method facilitates automatic parameter optimization in some cases and assists manual searches in others.
Conclusions:
- The developed method offers a robust approach for fitting connectionist models to empirical data.
- This tool enhances the accuracy and interpretability of computational cognitive models.
- The method is implemented in Matlab, documented, and freely available.