Optuna bayesian optimization starter
A framework for discovering, compiling, and validating reusable skills for scientific agents.
npx -y skills add ma-compbio-lab/SkillFoundry --skill optuna-bayesian-optimization-starterAssembled from the repository path, not quoted from the project. Check it against their README if it does not work.
SKILL.md
1.1 KB, 226 tokens by cl100k_base, as published. Nobody here has run it
Optuna Bayesian Optimization Starter
Use this skill to run a deterministic Optuna optimization loop on a small synthetic objective and inspect the best trial summary.
What This Skill Does
- defines a two-parameter toy objective
- optimizes it with Optuna's TPE sampler
- records the best trial, best parameters, and a short ranked trial table
When To Use It
- when you need a runnable
bayesian-optimizationstarter - when you want a local Optuna example before wiring in an expensive scientific objective
- when you need deterministic optimization outputs for repository tests
Run
./slurm/envs/statistics/bin/python skills/statistical-and-machine-learning-foundations-for-science/optuna-bayesian-optimization-starter/scripts/run_optuna_bayesian_optimization.py --out scratch/optuna/bayesian_optimization_summary.json
Notes
- The objective is synthetic and smooth on purpose; it exists to verify the optimization loop, not to benchmark samplers.
- Increase the trial budget only after replacing the toy objective with a real scientific function.