PyTorch
Safetensors
braindecode
deprecated
bruAristimunha commited on
Commit
a0bd100
·
verified ·
1 Parent(s): 821235f

Complete training-data provenance and Braindecode software citation

Browse files

README-only metadata audit, 2026-11-09. Braindecode Zenodo citation. No weights, config, license, or upstream paper citations changed.

Files changed (1) hide show
  1. README.md +35 -0
README.md CHANGED
@@ -54,3 +54,38 @@ accessions, Hub ids otherwise). `nm`/`on` ids are datasets on [NEMAR](https://ne
54
  - S-JEPA self-supervised pre-training (16 s examples, 60 % spatial mask, run adeuwv4s) on Lee2019 subjects 1-40, all three paradigms pooled, no labels; subjects 41-47 = SSL validation set used for early stopping / best-checkpoint selection; subjects 48-54 not used for these weights (downstream evaluation only). Only the pre-trained backbone tensors come from adeuwv4s (byte-identical); the SignalJEPA_PostLocal head (19 channels, 2 outputs, plus a 19-row spatial embedding for Contextual) has no trace of supervised training on any dataset (no fine-tuning is documented; the original README states the downstream head 'was not pre-trained').
55
  - Sources: arXiv:2403.11772 (HTML), section DATASETS; https://huggingface.co/api/models/braindecode/signal-jepa/commits/main (commit 9e20126cf5); braindecode check: tensor-byte sha256 check -> compare_sjepa.json; W/bd/braindecode/models/signal_jepa.py.
56
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
54
  - S-JEPA self-supervised pre-training (16 s examples, 60 % spatial mask, run adeuwv4s) on Lee2019 subjects 1-40, all three paradigms pooled, no labels; subjects 41-47 = SSL validation set used for early stopping / best-checkpoint selection; subjects 48-54 not used for these weights (downstream evaluation only). Only the pre-trained backbone tensors come from adeuwv4s (byte-identical); the SignalJEPA_PostLocal head (19 channels, 2 outputs, plus a 19-row spatial embedding for Contextual) has no trace of supervised training on any dataset (no fine-tuning is documented; the original README states the downstream head 'was not pre-trained').
55
  - Sources: arXiv:2403.11772 (HTML), section DATASETS; https://huggingface.co/api/models/braindecode/signal-jepa/commits/main (commit 9e20126cf5); braindecode check: tensor-byte sha256 check -> compare_sjepa.json; W/bd/braindecode/models/signal_jepa.py.
56
 
57
+
58
+
59
+ ## Cite Braindecode
60
+
61
+ If you use this braindecode port in scientific work, cite the software on Zenodo,
62
+ [doi:10.5281/zenodo.17699192](https://doi.org/10.5281/zenodo.17699192),
63
+ and the original model and datasets cited above.
64
+
65
+ ```bibtex
66
+ @software{braindecode,
67
+ author = {Aristimunha, Bruno and
68
+ Guetschel, Pierre and
69
+ Wimpff, Martin and
70
+ Gemein, Lukas and
71
+ Rommel, Cedric and
72
+ Banville, Hubert and
73
+ Sliwowski, Maciej and
74
+ Wilson, Daniel and
75
+ Brandt, Simon and
76
+ Gnassounou, Théo and
77
+ Paillard, Joseph and
78
+ {Junqueira Lopes}, Bruna and
79
+ Sedlar, Sara and
80
+ Moreau, Thomas and
81
+ Chevallier, Sylvain and
82
+ Gramfort, Alexandre and
83
+ Schirrmeister, Robin Tibor},
84
+ title = {Braindecode: toolbox for decoding raw electrophysiological brain data
85
+ with deep learning models},
86
+ url = {https://github.com/braindecode/braindecode},
87
+ doi = {10.5281/zenodo.17699192},
88
+ publisher = {Zenodo},
89
+ license = {BSD-3-Clause},
90
+ }
91
+ ```