/radd

Race Against Drift-Diffusion model of proactive and reactive response inhibition

Primary LanguagePython

RADD: Race Against Drift-Diffusion model of sensorimotor inhibition and decision-making

Summary

RADD is a python package for modeling the underlying dynamics of motor inhibition as a combination of two widely utilized conceptual frameworks: race models of response inhibtion and drift-diffusion models of decision-making.

RADD seeks to explain both proactive and reactive forms of response inhibition within a unified framework in which the competition between direct ("Go") and indirect ("No Go") pathways is modeled as a stochastic accumulation of evidence between "Respond" and "Inhibit" boundaries. This diffusion process acts as a dynamically moving baseline from which a hyperdirect "Stop" process can be initiated. In the event that a stop signal is encountered, the hyperdirect pathway must override the current level of "Go" evidence in order to suppress the evolving motor response.

Features

  • Includes models of proactive and reactive stopping.

  • Gradient descent optimization of drift-diffusion parameters.

  • Flexible control over parameter dependencies.

  • Include dynamic bias signal (see Hanks et al., 2011)

  • Visualizations for assessing go and stop RT distributions, comparing alternative model fits, overlaying simulated data on empirical means, etc.

  • Simulate neural integration of direct, indirect, and hyperdirect pathways in the Basal Ganglia - useful for generating and testing predictions about fMRI and single-unit electrophysiological data.