BRAID: Input-driven Nonlinear Dynamical Modeling of Neural-Behavioral Data
Authors: Parsa Vahidi, Omid G. Sani, Maryam Shanechi
ICLR 2025 | Venue PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We validate BRAID with nonlinear simulations, showing that it can accurately learn the intrinsic dynamics shared between neural and behavioral modalities. We then apply BRAID to motor cortical activity recorded during a motor task and demonstrate that our method more accurately fits the neural-behavioral data by incorporating measured sensory stimuli into the model and improves the forecasting of neural-behavioral data compared with various baseline methods, whether input-driven or not. |
| Researcher Affiliation | Academia | Parsa Vahidi1, Omid G. Sani1, Maryam M. Shanechi1-3,* 1Electrical and Computer Engineering, 2Computer Science, 3Biomedical Engineering Viterbi School of Engineering, University of Southern California (USC), Los Angeles, CA EMAIL, *Corresponding author |
| Pseudocode | No | The paper provides detailed mathematical formulations and descriptions of the learning algorithm steps in prose (e.g., Appendix A.1.2 'LEARNING ALGORITHM STEPS'), but it does not include an explicitly labeled 'Pseudocode' or 'Algorithm' block with a structured, code-like format. |
| Open Source Code | Yes | To ensure the reproducibility of our work, we are sharing the code for BRAID along with a Python notebook demonstrating its usage at https://github.com/ShanechiLab/BRAID. |
| Open Datasets | Yes | We then apply our method to electrophysiological data recorded from a non-human-primate (NHP) performing sequential reaches (O Doherty et al., 2017). ... Finally, the dataset we used (O Doherty et al., 2017) is publicly available for anyone interested in reproducing the results reported in section 4.2. |
| Dataset Splits | Yes | To evaluate the performance of our models, we perform 5- and 2-fold cross-validation, for real data and simulation analyses, respectively. |
| Hardware Specification | No | The paper does not provide specific hardware details (e.g., GPU model, CPU type, memory amounts, or specific cloud computing instances) used for running the experiments or training the models. |
| Software Dependencies | No | The paper mentions the use of an 'Adam optimizer (Kingma & Ba, 2017)' and 'sklearn.neighbors.KNeighborsRegressor (Pedregosa et al., 2011)' but does not provide specific version numbers for these or other general software libraries and frameworks used for the implementation of BRAID itself. |
| Experiment Setup | Yes | Table A.2: BRAID hyperparameters used in real data experiments and simulations Hyperparameter Value Number of hidden layers in nonlinear maps 1 Number of hidden units in nonlinear maps 64 Nonlinear activation ReLU Learning rate 0.001 Batch size 32 Sequence length 128 Optimizer Adam |