The Courtois project on neuronal modelling: scaling up AI models of individual brains in a massive individual fMRI dataset
Bibliographic record
Abstract
The Courtois project on neuronal modelling (CNeuroMod) aims to improve the task flexibility of artificial neural networks using activity recorded from biological neural networks. CNeuroMod has collected an unprecedented “deep'' functional magnetic resonance imaging (fMRI) dataset currently available, with up to 150 hours of fMRI data per subject (N=6) [1]. The project covers a wide range of tasks, broadly categorised in domains such as vision, audition, language, memory, emotions and videogames. In this talk, I will provide an overview of the contents of the Courtois NeuroMod database, and then present initial results on scaling up AI models of individual brains. I will first discuss some recent works on brain decoding using the human connectome project (HCP) task battery. We found a marked advantage of deep graph convolutional networks for group level decoding on that benchmark using the full original HCP sample (N=1200) [2,3]. Using Courtois Neuromod’s hcptrt dataset, where we repeated HCP’s task battery up to 15 times per subject, we successfully trained brain decoding models on single individuals that matched the performance of the group decoder, using two orders of magnitude less data [4]. I will also discuss a recent benchmark of auto-regressive models for individual video-watching fMRI data (i.e. movie10 and friends dataset), where we looked at over 10 different time series models, and up to 10 hours of training data [5]. We found that graph convolutional networks had the best performance overall, and performance scaled up with the amount of data without reaching a ceiling at 10 hours. Overall, these results demonstrate that large individual fMRI dataset can be used to efficiently train purely individual AI models of brain activity, and that massive amounts of individual data are beneficial to this endeavour. References [1] https://www.docs.cneuromod.ca [2] Zhang, Y., Tetrel, L., Thirion, B., Bellec, P., 2021. Functional annotation of human cognitive states using deep graph convolution. Neuroimage 231, 117847. https://doi.org/10.1016/j.neuroimage.2021.117847 [3] Zhang, Y., Farrugia, N., Bellec, P., 2022. Deep learning models of cognitive processes constrained by human brain connectomes. Med. Image Anal. 80, 102507. https://doi.org/10.1016/j.media.2022.102507 [4] Rastegarnia, S., Tetrel, L., Pinsard, B., DuPre, E., Zhang, Y., & Bellec, P. (2022, September 16). Brain decoding of the Human Connectome Project Tasks in a Dense Individual fMRI Dataset. Psyarxiv preprint https://doi.org/10.31234/osf.io/9t5nh [5] Paugam, F., Pinsard, B., Lajoie, G., & Bellec, P. (2023, February 17). A benchmark of individual auto-regressive models in a massive fMRI dataset. Psyarxiv preprint https://doi.org/10.31234/osf.io/pvx3d
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.002 | 0.009 |
| Meta-epidemiology (narrow) | 0.003 | 0.001 |
| Meta-epidemiology (broad) | 0.001 | 0.002 |
| Bibliometrics | 0.002 | 0.001 |
| Science and technology studies | 0.001 | 0.001 |
| Scholarly communication | 0.002 | 0.002 |
| Open science | 0.003 | 0.002 |
| Research integrity | 0.002 | 0.003 |
| Insufficient payload (model declined to judge) | 0.005 | 0.005 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".