bayesian-optimization
Guides modAL BayesianOptimizer and PI/EI/UCB acquisition workflows for expensive scalar-function optimization with GP-style regressors.
- License 许可
- MIT
- 最近更新
- 2026-09-03
- Stars
- ★ 351
- Forks
- 27
01 前言 Front Matter
- name
- bayesian-optimization
- license
- MIT
- metadata
- { "disco-role": "operating" }
- description
- Guides modAL BayesianOptimizer and PI/EI/UCB acquisition workflows for expensive scalar-function optimization with GP-style regressors.
- disable-model-invocation
- true
02 资源文件 4
- 参考文档references/acquisition-reference.md7.1 KB
- 参考文档references/troubleshooting.md5.8 KB
- 参考文档references/workflows.md6.6 KB
- 脚本scripts/bayesian_optimizer_smoke.py9.1 KB
03 SKILL.md 正文
Bayesian Optimization
Use this sub-skill when a task asks for Bayesian optimization with modAL: build a BayesianOptimizer, choose PI/EI/UCB acquisition behavior, run a bounded objective-evaluation loop, inspect acquisition scores, or debug get_max()/y_max tracking.
Read or run first
- Read references/workflows.md when you need an end-to-end optimizer loop, shape conventions, 1-D or multidimensional candidate grids, or evaluation-budget guidance.
- Read references/acquisition-reference.md when choosing among
max_PI,max_EI,max_UCBor inspectingoptimizer_PI,optimizer_EI,optimizer_UCBscore arrays. - Read references/troubleshooting.md when
teach,query, acquisition utilities,predict(return_std=True), orget_max()behave unexpectedly. - Run scripts/bayesian_optimizer_smoke.py to check that the installed
modALpackage can execute a deterministic Bayesian optimization loop and printPASSwithget_max()results.
Route boundaries
- For general
ActiveLearner,Committee,query,teach,fit,return_metrics, bagging, or estimator lifecycle questions, route to ../learners-and-committees/SKILL.md. - For generic query strategy selection, custom strategy combinators, uncertainty/disagreement sampling, ranked batch, density, expected-error, or multilabel strategies, route to ../query-strategies/SKILL.md.
- Keep this sub-skill focused on scalar-objective maximization with regressors that can provide both predictive mean and predictive standard deviation.
Minimal operating checklist
- Define a finite candidate pool
X_poolwith shape(n_candidates, n_features). - Seed
BayesianOptimizerwith at least one evaluated point:X_traininghas matching feature columns andy_trainingcontains numeric scalar objective values. - Use a regressor whose
predict(X, return_std=True)returns(mean, std);GaussianProcessRegressoris the usual modAL-backed choice. - Query before evaluating the expensive objective:
query_idx, query_inst = optimizer.query(X_pool). - Evaluate only
query_inst, normalize the result to one scalar per queried row, and calloptimizer.teach(query_inst, y_new). - Call
optimizer.get_max()for the best observed location/value so far; do not recompute the maximum from unevaluated candidate predictions.
04 同仓库更多 Skills
passnet-triton-opt
Make a single PassNet Triton kernel fast AND numerically faithful: performance model (when a replacement can win at all), block/grid/warp tuning, autotune policy, launch overhead, and per-op numeric recipes to pass the dtype baseline tolerances. Use when a pass matches and is correct but speedup below expected, or when correctness fails by small numeric margins.
passnet-solve
END-TO-END playbook for solving one PassNet sample: analyze the computation graphs, decide the optimization strategy (what to fuse, what to replace, what to leave alone), drive the iteration loop, and maximize the sample score. This is the ENTRY skill — invoke it first for any PassNet optimization task; it tells you when to use passnet-pattern-fusion, passnet-triton-opt and passnet-feedback.
passnet-skill
PassNet GPU kernel optimization via compiler passes. Design and implement Triton-based optimization passes, create pass files under ./pass_dir/, self-evaluate with pass_evaluator, and iterate to maximize GPU speedup.
passnet-pattern-fusion
Author PassNet pass files: write patterns that actually MATCH the FX graph, pick fusion regions (which ops to absorb into one Triton kernel), and structure multi-pass submissions with the mandatory shared-dispatch architecture. Use when creating or fixing pattern()/replacement_args()/replacement_func() files, when a pass "failed to match", or when deciding how to fuse multiple kernels/ops.
passnet-orchestrate
Round/batch planning and decision support for PassNet graph optimization. Use for multi-sample triage, worker allocation, eval-budget policy, and post-eval keep/revert decisions. For solving one concrete sample end to end, use passnet-solve as the entry skill; consult this skill only when you need broader planning or a specific decision gate.
passnet-feedback
Fast, deterministic feedback for PassNet work: pre-flight pattern/match verification WITHOUT burning a GPU evaluation, per-node bottleneck analysis of a sample's graphs, and eval-log parsing into per-variant status + estimated score + failure classification. Use BEFORE every GPU evaluation (check_pattern), at the START of a sample (analyze_graph), and AFTER every evaluation (parse_eval_log).