Bayesian modeling of structural connectivity #20
Description
Title
Bayesian modeling of structural connectivity
Short description and the goals for the OHBM BrainHack
The goal with this project is to develop an atlas of the reliability of structural connectivity matrices based on Bayesian modeling.
That model should fulfill the following two objectives: 1) quantify the reliability of fiber density estimation for each connection, 2) obtain a probability indicating whether the connection is most likely absent or present. I have developed a first model implemented in the Python library PyMC, but this model needs improvement.
Goals:
- Modify the current model to generate a good estimate of the probability that a connection is truly absent or present, based on simulated repeated measurements of a structural connectivity matrix. Possible leads include regularization, modify parameter prior distribution and modify model implementation.
- Incorporate connection length in the model
- Fit that model on real structural connectivity matrices and interpret the results
See this abstract for more details about the current status of the project: https://www.overleaf.com/read/zhsnjpskpzzh#d2befe
Link to the Project
https://github.com/TheAxonLab/hcph-sops
Image/Logo for the OHBM brainhack website
Project lead
Céline Provins, Github: celprov, Discord: cprovins
Main Hub
Brisbane
Link to the Project pitch
No response
Other hubs covered by the leaders
- Brisbane
- Hybrid (Asia / Pacific)
- Hybrid (Europe / Middle East / Africa)
- Hybrid (Americas)
Skills
- People with good knowledge of Bayesian modeling and probabilistic programming would be gladly appreciated, but if you're willing to learn more about Bayesian modeling with little to no previous experience with Bayesian statistics, you're more than welcome as well.
- confirmed level in one coding language (e.g Python or R) is preferred
- Git level 2: comfortable working with branches and can do a pull request on another repository is preferred
- If you're not comfortable with git or/and coding, but you think you can help with Bayesian statistics, please join
- If someone feels like they could particularly help us with their knowledge about structural connectivity, feel free to join for brainstorming
Recommended tutorials for new contributors
https://www.pymc.io/projects/docs/en/latest/learn/core_notebooks/pymc_overview.html
https://nbviewer.org/github/CamDavidsonPilon/Probabilistic-Programming-and-Bayesian-Methods-for-Hackers/blob/master/Chapter2_MorePyMC/Ch2_MorePyMC_PyMC3.ipynb
https://nbviewer.org/github/CamDavidsonPilon/Probabilistic-Programming-and-Bayesian-Methods-for-Hackers/blob/master/Chapter3_MCMC/Ch3_IntroMCMC_PyMC3.ipynb
Good first issues
- Re-run the Bayesian model, but learning one of the parameters of the Bayesian model instead of fixing it
- Re-run the Bayesian mode,l but using the same parameter for the Gaussian and Exponential part of the model
Twitter summary
No response
Short name for the Discord chat channel (~15 chars)
bayesian-sc
Please read and follow the OHBM Code of Conduct
- I agree to follow the OHBM Code of Conduct during the hackathon