This organization regroups works/packages/toolboxes related to performance estimation problems.
If you find some content useful, please don't hesitate to give feedbacks and or to star the content!
- PEPit: allows a quick access to performance estimation problems in Python.
- PEPit.jl: Julia version.
- PESTO: heritage Matlab version.
There are also related works on using performance estimation problems to find Lyapunov (or potential or energy) functions. Those techniques are not yet included in PEPit and PESTO (for structural reasons), which currently only allow studying predefined Lyapunov functions (rather than finding them). Three of those works provide easy-to-use interfaces:
- historical reference (outdated matlab code): Lyapunov functions for linear convergence of first-order methods for smooth strongly convex problems,
- AutoLyap (in Python).
- AutoLyap.jl (in Julia).
- Informal introduction to PEPs: here.
- New a complete tutorial on performance estimation problems with interactive labs here
- Older (shorter) tutorial with exercises: here.
If you organize a PEP-event, we would be happy to list your event below.
Past events:
- March 2026: Interactive mini-course (9h) at SMAI-MODE
- February 2023: PEP-talks.
If you have any feedback or other input, please don't hesitate to contribute by sharing it. We are happy to review any constructive pull request. In particular, if you find any example of application that you find relevant, don't hesitate to submit it as a new example within one of the toolboxes and/or to the tutorials:
- heritage learning performance estimation problems repository, or
- brand-new and complete 9h mini-course on performance estimation.
