HSSM: A Widely Applicable Toolbox for Hierarchical Bayesian Neurocognitive Modeling

Area of Science:

• Cognitive Neuroscience
• Computational Neuroscience
• Bayesian Inference

Background:

• Computational models are crucial in cognitive neuroscience but often limited to simple models due to analytical constraints.
• Rigorous application of complex models to experimental data is challenging.

Purpose of the Study:

• Introduce the Hierarchical Sequential Sampling Model (HSSM) ecosystem, a Python toolbox.
• Democratize access to a wide range of neurocognitive process models.
• Facilitate rigorous empirical testing of computational models.

Main Methods:

• Utilize hierarchical Bayesian inference and simulation-based inference with likelihood surrogates.
• Employ PyMC and Bambi for user-friendly formula syntax in hierarchical mixed-effects regressions.
• Incorporate trial-by-trial neural or physiological covariates.

Main Results:

• Enable fast parameter estimation for models lacking closed-form likelihoods.
• Facilitate rapid model simulation and training data generation.
• Provide utilities for training neural networks to deploy surrogate likelihoods.

Conclusions:

• The HSSM ecosystem accelerates the development and empirical testing cycle for computational models.
• Bridge the gap between computational theorists and experimentalists in cognitive neuroscience.
• Foster community-wide benefits through accessible and extensible modeling tools.