Beyond Drift Diffusion Models: Fitting a Broad Class of Decision and Reinforcement Learning Models with HDDM

Area of Science:

• Cognitive Neuroscience
• Computational Neuroscience
• Decision Science

Background:

• Computational modeling is crucial in cognitive neurosciences for understanding decision-making.
• Traditional analysis of sequential sampling models (SSMs) is limited to analytically tractable models.
• Integrating SSMs with reinforcement learning models is a recent development.

Purpose of the Study:

• To extend the HDDM Python toolbox for fitting and assessing a wider range of SSMs.
• To facilitate the combination of SSMs with reinforcement learning models.
• To provide users with tools for model visualization and assessment.

Main Methods:

• Utilizing likelihood-free inference methods for fitting complex SSMs.
• Integrating sequential sampling models with reinforcement learning algorithms.
• Developing extensions for the HDDM Python toolbox.

Main Results:

• The enhanced HDDM toolbox allows for the assessment of a broader spectrum of SSMs.
• Users can now combine SSMs with reinforcement learning models within a unified framework.
• The toolbox includes features for model visualization and posterior predictive checks.

Conclusions:

• The extended HDDM toolbox significantly advances the computational modeling capabilities in cognitive neuroscience.
• This work enables more sophisticated analyses of decision-making by integrating diverse modeling approaches.
• The user-friendly interface and comprehensive tools empower researchers to explore complex cognitive and neural processes.