cloelib and cloelike Posterior Sampling

[gcanasherrera]

💬 Discussion: cloelib and cloelike Posterior Sampling — Problems, Suggestions & Ideas

Hi everyone 👋

This thread is meant to centralize questions, problems, and suggestions related to posterior sampling with cloelib and cloelike.
The goal is to share experience, debug issues, and brainstorm improvements to make the sampling workflow more robust, efficient, and user-friendly.

---

🧩 Context

cloelib and cloelike are designed to provide flexible interfaces for connecting theoretical predictions and likelihood evaluations.
As they are increasingly used for posterior sampling, several recurring topics have emerged:

• Standardizing the interface between likelihoods and samplers
• Managing priors, parameter transformations, and derived quantities
• Improving performance and convergence in high-dimensional spaces
• Parallelization and other issues when running in clusters

---

⚙️ Current Issues and Discussion Points

Please feel free to expand on these or add your own:

• Performance bottlenecks in large or complex parameter spaces
• Consistency and reproducibility between cloelib and cloelike posteriors
• MPI vs. OpenMPI
• Suggestions for diagnostics or convergence tools

---

💡 How to Contribute

If you’ve:

• encountered sampling issues or strange posterior behavior,
• found optimizations or useful tricks,
• implemented new interfaces or improvements, or
• have ideas for expanding functionality —

please share them here!
Snippets, performance benchmarks, conceptual ideas, or test results are all welcome.
This will help align ongoing developments and guide the roadmap for both cloelib and cloelike.

---