Using the NVIDIA-CUDA framework, mcmc_min is able to efficiently sample the Bayesian posterior distribution over the parameters of a physiologically derived model of a task-switching and distractor inhibition paradigm. The model features a working-memory module, implementing the currently active task rule in terms of a two-dimensional stochastic dynamical system with three attractor states (rule 1, rule 2, spontaneous). The working memory module is given inputs representing the task cue in the simulated paradigm and the activity of the rule 1 versus rule 2 representing population biases a decision module. The decision module implements a three dimensional nonlinear drift-diffusion process derived from a physiological winner-take-all network with four selective populations, representing the possible behavioral decisions. The decision module integrates top-down input from the rule module with bottom-up input from the decision module to produce decisions and reaction time distributions.
Cognitive Stability and Flexibility FIT
GPU Accelerated Fitting of Behavioral Data by a Physiological Model
Status: Beta
Brought to you by:
kueltzho
Downloads:
2 This Week