Repository navigation
Description
Task
T-055b· Milestone M5 — Intelligent selection · Size XS · Priority P1 · Epic #5Implements NFR-PERF-002 · NFR-SCAL-001
Branch
feat/t-055b-robustnessgp-scale-guard-informative-subsampling— one task = one branch = one PR (doc 32 §6)
Summary
When n > selection.gp.max_points, train on the most informative subset (all failures, all |ρ| < 0.2, then most recent) and log gp.subsampled=true.
Context / motivation
Exact GP is O(n³); campaigns of 10 000 runs (NFR-SCAL-001) need a guard to keep propose() within budget.
Scope
In scope
select_training_subset(X, y, max_points, order)pure function + integration infit.
Implementation guidance
- Make the
|ρ| < 0.2band a settings key (selection.gp.boundary_band, new) — no magic numbers. - Standards: respect the layering contracts (doc 16 §2) and conventions in doc 34; no hard-coded tunables — thresholds/defaults come from typed settings (doc 27); no magic numbers (ruff
PLR2004); public APIs typed (mypy strict) and documented.
Configuration keys
From the configuration reference (doc 27). Keys marked (new) must be added to the typed settings models and appear in the generated reference (T-003c).
selection.gp.max_points
Acceptance criteria
- Subset size ≤ max_points; contains all failures when they fit.
- Fit time at n=5 000 bounded (benchmark recorded).
- Tests listed below added and green in CI; coverage of new code meets the gate in doc 18 §3.
- Docs, config reference and CHANGELOG
Unreleasedupdated where behaviour, keys or interfaces changed. - Definition of Done (doc 32 §3) satisfied.
Tests required
tests/unit/selection/test_subsampling.py.
Dependencies
Blocked by: #110 (T-055a)
Blocks: no other task (leaf)
Parallelization notes
Files / paths owned by this task — avoid editing other paths; shared files (pyproject.toml entry points, cli/main.py, api/routers/__init__.py, router.tsx, docker/compose.yaml, ci.yml) are append-only:
- backend/src/roboprobe/selection/surrogate.py (subset fn)
- backend/src/roboprobe/core/settings/selection.py (1 field)
Small.
Definition of Done
Per doc 32 §3 and AGENTS.md §5: lint, format, mypy strict and import-linter green; tests added and passing; docs updated; PR template completed with Closes #<this issue>; Conventional Commit messages referencing T-055b (doc 16 §6); commits authored only by the repository owner identity (AGENTS.md §4); CI green; reviewer checklist (doc 32 §5.1) satisfied.
References
Design docs: doc 07 §4 · doc 07 §6.4
Requirements: NFR-PERF-002 · NFR-SCAL-001 (traceability: doc 28)
Metadata
Metadata
Assignees
Labels
Projects
- StatusShow more project fieldsBacklog
Activity