Fuzzy Pipelines

[gkiar]

Project Summary

Field	Value
Project Name	Fuzzy Pipelines
Description	How much can we trust our pipelines to give us meaningful answers? One way to answer this is to introduce tiny perturbations throughout our pipelines, and see how much our answers change. This project aims at making it easier to perform these analyses and answer these questions.
URL(s)	Basic math issues notebook 😨, Stability Paper, Fuzzy Environments repo, Verificarlo repo
Contributors' Github Slack handle	@gkiar, @glatard
Image

Will teach...

• [x] Docker
• [x] Numerical Analysis/Monte Carlo Arithmetic
• [x] Continuous Integration
• [x] Testing

Need help with...

• [ ] Write a Dockerfile for your tool that uses the fuzzy ecosystem
• [ ] Write a CI build script that runs + evaluates fuzzy arithmetic for built containers
• [ ] Turn the tutorial notebooks into a Jupyter book
• [ ] Close any open issue on the fuzzy repository
• [ ] Create cartoon logo of a marimo and a division symbol (or other math + fuzzy combinations!)

[bpinsard]

Does it only focus on numerical instabilities or on measurement noise as well? For instance adding random Rician noise to MRI input and evaluate the stability of result.

[gkiar]

@bpinsard Both! An important point is understanding the relationship between numerical instabilities which we can simulate using data-agnostic techniques like those above, and those from more structured data-specific noise. Rician, of course, at the top of the list for MRI 😄

[gkiar]

Update from the hackathon: https://github.com/gkiar/fuzzy/issues/5