Identifying Connectivity Distributions from Neural Dynamics Using Flows
Paper • 2603.26506 • Published
Dataset Viewer
The dataset viewer is not available because its heuristics could not detect any supported data files. You can try uploading some data files, or configuring the data files location manually.
YAML Metadata Warning:empty or missing yaml metadata in repo card
Check out the documentation for more information.
Connector reproduction bundle
This bundle reproduces Claims 1–4 of arXiv:2603.26506v2. Claim 5 is not independently rerunnable because the released materials omit the trained FINDR model/loadings and distilled lrRNN used for the representative rat session.
Create an environment with PyTorch, NumPy, SciPy, scikit-learn, and
flow-matching==1.0.10, then run from the bundle root:
python scripts/c1_flow_distribution.py \
--checkpoint outputs/local_quadstable/model_fold_1_seed_1_id_0.pth \
--data data/quadstable_slim.npz --source connector_src \
--output-dir outputs/claim1 --samples 5000
python scripts/c2_linear_identifiability.py
python scripts/c3_rank_conditions.py
python scripts/c4_maxent.py
The paper-scale Hugging Face training/evaluation evidence is recorded at:
- Training logs: https://huggingface.co/jobs/GwendalTsang/6a5e123dbee6ee1cf4ed250b
- Completed evaluation: https://huggingface.co/jobs/GwendalTsang/6a5e13e6d216bd6f3a20364f
- Persisted checkpoint and metrics: https://huggingface.co/buckets/GwendalTsang/repro-connector-runs
Official method source: https://github.com/AllenInstitute/connector
- Downloads last month
- 131