Modeling state-dependent communication between brain regions with switching nonlinear dynamical systems

Authors: Orren Karniol-Tambour, David M. Zoltowski, E. Mika Diamanti, Lucas Pinto, Carlos D Brody, David W. Tank, Jonathan W. Pillow

ICLR 2024 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We show that our model accurately recovers latent trajectories, vector fields underlying switching nonlinear dynamics, and cross-region communication profiles in three simulations. We then apply our method to two large-scale, multi-region neural datasets involving mouse decision-making.
Researcher Affiliation Academia Orren Karniol-Tambour, David M. Zoltowski, E. Mika Diamanti, Lucas Pinto , Carlos D. Brody, David W. Tank, Jonathan W. Pillow Princeton Neuroscience Institute, Northwestern University {orrenkt,dzoltowski,mdiamanti,cbrody,dwtank,jpilllow}@princeton.edu lucas.pinto@northwestern.edu
Pseudocode Yes Algorithm Cosmoothing multiregion drop out training
Open Source Code No The paper mentions using a modified version of the SSM package from Linderman et al. (https://github.com/slinderman/ssm), but it does not provide a link or statement about open-sourcing the code for MR-SDS, the methodology described in this paper.
Open Datasets Yes We applied our method to calcium imaging data recorded in mice performing a sensory evidence accumulation task (Figure 4). In the task, a headfixed mouse runs on a linear track in virtual reality while columns ( towers ) are presented on both sides of the track (Pinto et al., 2019).
Dataset Splits Yes The Neural Latents Benchmark uses a 25% neuron drop out rate, and we apply this rate, with one modification we evaluate all models against a 25% drop out rate per region.
Hardware Specification Yes All experiments and analysis were run on a 28 CPU, 8 GPU (Ge Force RTX 2080 Ti) server.
Software Dependencies No The paper mentions using 'a modified version of the SSM package' but does not specify version numbers for this or any other software components used for the experiments.
Experiment Setup Yes Table 4: Parameters used for each of the experiments presented in the paper. LV: Switching Lotka-Voltera. DW: Double well. Meso: mesoscope. Wide: widefield. lr: learning rate. p2: trial dropout rate used for cosmoothing training. I: Identity mapping. L: linear layer. d: dimensionality of latents for each region.