Learning Time-Varying Multi-Region Brain Communications via Scalable Markovian Gaussian Processes

Authors: Weihan Li, Yule Wang, Chengrui Li, Anqi Wu

ICML 2025 | Venue PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We validate our model using neural recordings from multiple regions of the brain during visual processing tasks (Semedo et al., 2019; Siegle et al., 2021). Our results demonstrate the method s capability to uncover how information flow patterns dynamically change across multi-region networks, offering new insights into the temporal organization of largescale neural circuits and advancing our understanding of distributed neural computation. We evaluate our model on three datasets.
Researcher Affiliation Academia 1School of Computational Science & Engineering, Georgia Institute of Technology, Atlanta, USA. Correspondence to: Anqi Wu <EMAIL>.
Pseudocode No The paper describes the Kalman EM algorithm and its steps in Appendix B, but it does not present it in a pseudocode block or a clearly labeled algorithm format. It uses mathematical equations and explanatory text.
Open Source Code Yes Code is available at https://github.com/ BRAINML-GT/Adaptive-Delay-Model.
Open Datasets Yes Two Brain Regions (Semedo et al., 2019; Zandvakili & Kohn, 2019): Simultaneous spike train recordings from a monkey s primary visual area (V1) and secondary visual cortex (V2)... Five Brain Regions (Siegle et al., 2021): Simultaneous spike train recordings from a mouse s primary visual cortex (VISp)...
Dataset Splits Yes We evaluate our model and baseline models by randomly splitting the data into training, validation, and testing sets with a ratio of 0.8, 0.1, and 0.1, respectively.
Hardware Specification No The paper mentions 'GPU parallelization' and 'modern hardware' but does not specify any particular GPU models, CPU models, or other detailed hardware specifications used for experiments.
Software Dependencies No The paper mentions 'MATLAB' in the context of a baseline (m DLAG) but does not provide specific version numbers for its own implementation's software dependencies.
Experiment Setup Yes The number of across-region and within-region latent dynamics follows previous works (Gokcen et al., 2022; Li et al., 2024b), where ma = 2 and mw = 2. The order P = 4 is selected based on performance evaluation on the validation dataset. ... we set ma = 2, mw = 1, and P = 5. ... we then conduct a grid search with 5-fold cross-validation to refine the number of across-region and within-region latent dynamics and the model order P.