All projects
NeuroDataReHack 2026

Brain-wide neural attractor dynamics of decision-making

Estrella Villanueva Pivel · Siyu Wang

Project Description

Previous evidence hints at attractor dynamics underlying decision-making processes. Despite of a large body of theoretical work, and the few empirical work on reconstructing attractor landscapes from neurophysiology data in single brain regions (e.g., PFC[1], PMD[2]), it is not understood how attractor dynamics is organized across different hierarchies of the brain, during decision-making. The aim of the project is to leverage the IBL - Brain Wide Map Dandiset [3] to study the brain-wide organization of attractor neural correlates, and how these attractor landscapes are modulated by external stimulus input versus internal state/prior.

Objectives and Approach

  • Examine if/how sensory stimulus intensity shifts the steepness of the attractor basin
  • Determine whether disengaged attentional states present a shallower attractor landscape
  • Future work - study how dynamics are coordinated across regions and hierarchies

Progress and Next Steps

  • Trained a low-rank RNN on the IBL task
  • Compared the 1-dimensional latent trajectories along the choice-encoding axis under different stimulus contrast levels
  • Computed and compared residual correlations, which reflected strength of attractor dynamics, in different stimulus contrast levels.
  • Fit a GLM-HMM-T to classify the behavior into engaged and disengaged states
  • Compared the choice separability under each of these states

Background and References

[1] Wang et al 2023, [2] Gerkin et al 2025, [3] IBL et al 2025, [4] Mohammadi et al 2025 , [5] Oña-Jodar et al 2024

Slides