Outer search now ranks individuals by (-n_fails, fitness) instead of raw
fitness scalar. This prevents high-score 3-fail designs from displacing
2-fail designs in tournament selection and population replacement — the
root cause of the §4.8 pathology where flag count dominates geometry.
Inner loop is unchanged: it still optimises against the raw 0.5^n fitness
scalar, so the cliff that prevents trading into new failures remains intact
(0/9 regressions in experiments/penalty_reshape.py).
Also removes stale _CHILD_INNER_KW = {"sigmas": (0.05,)}: this was left
over from the CMA-ES era; the NM inner loop default (homemaker-py-d6d)
does not accept a sigmas parameter.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
programme-house budget=20000: 1.04e-02 (2 fails), 1.36× over Phase-2
oracle run and 2.60× over urb-evolve p128. Winning topology found via
rotate at eval 10357, unreachable within Phase-2 budget. 71.8 evals/s
(~140× faster than batched oracle).
harbor-house (16 rooms): 3.73e-18 (49 fails) at budget 10000 in 633s.
This programme is beyond the oracle's capability; native fitness makes
it feasible. 638 topologies explored.
Adds experiments/run_search_scaled.py (native-only search runner, no
oracle dependency). DESIGN.md records Phase 3 gate result.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
benchmark_vs_urbevolve.py results (2026-06-13, budget=2000, URB_NO_OCCLUSION=1):
- Seeded designs: memetic beats urb-evolve 1.91× (c964435) and 1.63× (2f45907)
- Blank slate init.dom: memetic at 18 fails vs urb-evolve at 6 fails (topology
diversity gap from single-seed mutation chain vs random-population init)
Bug fixed: run_search.py was calling oracle.score on out.parent without
patterns.config present — causing the re-score to return near-zero instead of
the correct tracked fitness. Added shutil.copy to propagate patterns.config
alongside the output .dom before the standalone re-score.
Gate recorded in DESIGN.md §7. Closes homemaker-py-way.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
4-way comparison (NM / CMA-ES / compass / compass-ms) over 3 corpus files ×
3 seeds at budget 200, cold-start, URB_NO_OCCLUSION=1. CMA-ES wins on
batch-efficiency (18 oracle calls vs 200 for NM, 12x speedup on Perl startup
amortisation per §4.6) with acceptable quality (x1.41 @200 vs NM's x1.56).
Compass stalls on narrow-valley landscapes and introduces fail regressions.
NM flagged as Phase 3+ candidate once native fitness removes oracle call overhead.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Corpus: all 35 scores shift (x1.0-1.24, daylight pinned), one expected
failure-set change (458aa8b8 +2 crinkliness), oracle ~8% faster batched.
New deterministic-seed reference gains become the accept_innerloop bars:
x1.63 / x1.70 / x1.68 at budget 400, ~35 oracle calls per topology.
urb-evolve respects the flag by construction (in-process fitness reads
ENV at call time). Old flag-off numbers kept in DESIGN as historical.
Closes homemaker-py-gp2 (Urb-side patch lives in /home/bruno/src/urb,
uncommitted there pending review).
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Strategy decision (Bruno): occlusion is orthogonal to building a scalable
optimiser. Instead of porting Sun/Occlusion/CIESky, disable it in Urb
behind an env flag (daylight -> 1, illumination factor -> 1 so crinkliness
is unweighted external wall area / floor area). Python occlusion rebuild
deferred until optimisation is fully native.
Tracker: new homemaker-py-gp2 (flag + re-baseline) gates gnw/way/uxz;
homemaker-py-2g5 re-scoped to the post-Phase-5 Python rebuild (P4).
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Self-contained design + plan structured for breaking into bd (beads) tasks:
domain constraints that fix the slicing representation; what was built;
full empirical record (geometry port validated 35/35; area-proxy solver
falsified; perpendicular artifact resolved via equal-offset cuts; full-fitness
frozen-topology optimisation validated with 24-67% headroom; 0.5^n cliff);
validated memetic architecture; component plan; phased roadmap; risks/open
questions; repro steps; gotchas.
Oracle throughput measured: ~0.99s/dom batched vs 1.65s single (assessment-
dominated). urb-fitness.pl batches many doms per call, so native fitness is a
later speed/scale optimisation, not a gate; favour population/batch optimisers
and prototype the search on the batched oracle first.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
