diff --git a/.beads/issues.jsonl b/.beads/issues.jsonl index d8bd191..f083518 100644 --- a/.beads/issues.jsonl +++ b/.beads/issues.jsonl @@ -22,7 +22,7 @@ {"id":"homemaker-py-1p0","title":"Geometry inner loop: full-objective equal-offset ratio optimiser","description":"DESIGN.md §5.1, §7 Phase 1. Productionise experiments/optimize_fullfitness.py into homemaker: optimise(topology, x0=None) -\u003e (geometry, fitness). DOF = equal-offset division ratios of free branches (solver.free_branches, lowest-storey cut ownership), clipped to [eps, 1-eps]. Objective = full oracle fitness (never a proxy — §4.2 falsified). Must support warm-start x0 (§5.6) and a population/batch evaluation mode so each iteration scores via one batched oracle call (§4.6).","acceptance_criteria":"Reproduces or exceeds §4.5 gains (x1.24–x1.67, no new failures) on 2f45907, candidate-002, c964435; works as a library call on any corpus .dom","status":"closed","priority":1,"issue_type":"feature","assignee":"Bruno Postle","owner":"bruno@postle.net","created_at":"2026-06-11T23:36:58Z","created_by":"Bruno Postle","updated_at":"2026-06-12T08:46:31Z","started_at":"2026-06-12T00:14:19Z","closed_at":"2026-06-12T08:46:31Z","close_reason":"innerloop.optimise() lands: batched CMA-ES sigma ladder (0.05/0.15, IPOP popsize doubling, deterministic seeding) over equal-offset free-branch ratios vs full oracle fitness; warm-start x0 supported. Acceptance vs unprojected originals: x1.65/x1.66/x1.58 against bars x1.24/x1.67/x1.59, no new failures, 46 oracle calls vs NM's 200. Two near-bar results accepted as reproduced-within-noise (1% tol) — draw spread brackets the single-NM-draw bars; approved by Bruno 2026-06-12. Gotchas: equal-offset projection of legacy unequal cuts loses fitness/adds failures (midpoint projection used); pycma seed=0 means clock-seeded.","dependencies":[{"issue_id":"homemaker-py-1p0","depends_on_id":"homemaker-py-av5","type":"blocks","created_at":"2026-06-12T00:39:33Z","created_by":"Bruno Postle","metadata":"{}"}],"dependency_count":1,"dependent_count":3,"comment_count":0} {"id":"homemaker-py-8cs","title":"Experiment: warm-vs-cold start of inner loop (Lamarckian inheritance)","description":"DESIGN.md §5.6, §4.6. Warm-starting a child topology's inner loop from the parent's optimised ratios is the main lever for cutting per-topology cost (~3 min/topology cold). Apply single topology mutations to optimised corpus designs, re-optimise warm (surviving cuts keep values, new cuts get heuristic defaults) vs cold, compare oracle-call counts to convergence at equal final fitness.","acceptance_criteria":"Speedup factor measured across \u003e=10 mutated topologies; decision recorded (expect order-of-magnitude; if \u003c2x, revisit §4.6 Phase-2 scoping)","notes":"Experiment script committed (experiments/warm_vs_cold.py, 1cc86c8) and machinery validated oracle-free; one mutated child scored through the oracle OK. Waiting on homemaker-py-gp2 reference run to finish, then execute under URB_NO_OCCLUSION=1 (3 parents x 400 evals + 12 children x 2 x 200 evals, ~1.5-2 h oracle time). Default budgets: parent 400, child 200; target = evals to 95% of best final.","status":"closed","priority":1,"issue_type":"task","owner":"bruno@postle.net","created_at":"2026-06-11T23:36:58Z","created_by":"Bruno Postle","updated_at":"2026-06-12T11:44:45Z","closed_at":"2026-06-12T11:44:45Z","close_reason":"Measured (URB_NO_OCCLUSION=1, parent budget 400, child 200, 12 single mutations across 3 designs): cold start reached 95% of warm final in 0/12 cases within budget — speedup unbounded at practical budgets; warm finals beat cold finals x1.2-x4 in 12/12; 6/12 warm starts were within 95% at 1 eval (near-neutral mutations). Decision: Lamarckian warm-starting is MANDATORY in the memetic driver (homemaker-py-b39), not an optimisation; cold starts produce strictly worse geometry at equal budget. Note: 2 undivides were exactly fitness-neutral (same-type merge == Merge_Divided equivalence) — locality datum for homemaker-py-nyb.","dependencies":[{"issue_id":"homemaker-py-8cs","depends_on_id":"homemaker-py-1p0","type":"blocks","created_at":"2026-06-12T00:39:34Z","created_by":"Bruno Postle","metadata":"{}"}],"dependency_count":1,"dependent_count":0,"comment_count":0} {"id":"homemaker-py-av5","title":"Batched oracle: score many .dom files per invocation","description":"oracle.py currently scores one .dom per urb-fitness.pl call (~1.65 s/dom). DESIGN.md §4.6: batching amortises Perl startup to ~0.99 s/dom and is required so population/batch optimisers can score a whole generation in one oracle call. Extend oracle.py with a batch API: write N .dom files, one perl invocation, parse N .score/.fails pairs. Keep the single-file path for compatibility.","acceptance_criteria":"Batch of 35 corpus files scores in one perl invocation; per-file results identical to single-file calls; measured s/dom reported","status":"closed","priority":1,"issue_type":"task","assignee":"Bruno Postle","owner":"bruno@postle.net","created_at":"2026-06-11T23:36:56Z","created_by":"Bruno Postle","updated_at":"2026-06-12T00:14:06Z","started_at":"2026-06-11T23:50:40Z","closed_at":"2026-06-12T00:14:06Z","close_reason":"score_batch() lands in oracle.py; 35-file corpus parity verified single-vs-batch (1e-12 rel fitness, exact fail sets); 0.98 s/dom batched vs 1.27 single, x1.30","dependency_count":0,"dependent_count":1,"comment_count":0} -{"id":"homemaker-py-9wi","title":"Adjacency-aware discrete assignment for finish-time collapse (QAP/CP-SAT)","description":"fitness.collapse_superposition (homemaker-py-9o5/94g family) already does exact optimal relabelling of superposed leaves via brute-force permutation (CLASS_CAP\u003c=4) or Hungarian (linear_sum_assignment) beyond that -- but _best_assignment's docstring is explicit that the objective is deliberately SEPARABLE per leaf (quality_size * quality_width * quality_proportion only); perpendicular/crinkliness/access/adjacency are assumed usage-invariant within a class and left out, because adjacency quality depends on PAIRS of leaf-label assignments, not one leaf at a time, which breaks the exact separable solve.\n\nProposal: extend the collapse step to account for adjacency between candidate labels -- either a quadratic-assignment-style local search (2-opt swaps over the current Hungarian solution, accepting swaps that improve total adjacency satisfaction) or a CP-SAT (OR-Tools) encoding of the labelling problem with pairwise adjacency terms. This directly extends the one search-adjacent technique (exact/near-exact discrete assignment) that has actually paid off in this project, into territory the current separable solve cannot reach.\n\nMeasure against the current Hungarian-only collapse on harbor-house (heavy interchange-class usage: neighborhoods, meeting rooms, individual rooms) where adjacency-blind relabelling is most likely to leave adjacency fails on the table.","status":"open","priority":2,"issue_type":"task","owner":"bruno@postle.net","created_at":"2026-07-25T20:18:24Z","created_by":"Bruno Postle","updated_at":"2026-07-25T20:18:24Z","dependency_count":0,"dependent_count":0,"comment_count":0} +{"id":"homemaker-py-9wi","title":"Adjacency-aware discrete assignment for finish-time collapse (QAP/CP-SAT)","description":"fitness.collapse_superposition (homemaker-py-9o5/94g family) already does exact optimal relabelling of superposed leaves via brute-force permutation (CLASS_CAP\u003c=4) or Hungarian (linear_sum_assignment) beyond that -- but _best_assignment's docstring is explicit that the objective is deliberately SEPARABLE per leaf (quality_size * quality_width * quality_proportion only); perpendicular/crinkliness/access/adjacency are assumed usage-invariant within a class and left out, because adjacency quality depends on PAIRS of leaf-label assignments, not one leaf at a time, which breaks the exact separable solve.\n\nProposal: extend the collapse step to account for adjacency between candidate labels -- either a quadratic-assignment-style local search (2-opt swaps over the current Hungarian solution, accepting swaps that improve total adjacency satisfaction) or a CP-SAT (OR-Tools) encoding of the labelling problem with pairwise adjacency terms. This directly extends the one search-adjacent technique (exact/near-exact discrete assignment) that has actually paid off in this project, into territory the current separable solve cannot reach.\n\nMeasure against the current Hungarian-only collapse on harbor-house (heavy interchange-class usage: neighborhoods, meeting rooms, individual rooms) where adjacency-blind relabelling is most likely to leave adjacency fails on the table.","status":"in_progress","priority":2,"issue_type":"task","assignee":"Bruno Postle","owner":"bruno@postle.net","created_at":"2026-07-25T20:18:24Z","created_by":"Bruno Postle","updated_at":"2026-07-26T14:50:36Z","started_at":"2026-07-26T14:50:36Z","dependency_count":0,"dependent_count":0,"comment_count":0} {"id":"homemaker-py-f1d","title":"Ruin-and-recreate LNS: rebuild wings with the adjacency-aware constructor mid-search","description":"DESIGN.md's own experiment log shows every 'search machinery' change (niching+restarts 11.5, graded objective 11.4, Wong-Liu reassociation+shape-feasibility 12.3, granularity 12.4, island model 14, grain annealing 16, circulation-repair ops 21/22) has come back null-to-negative, while construction/seeding quality (adjacency-aware seeding 11.6/11.7, proportion-aware seeding 12.2) is the only lever that has ever moved the fail count. operators._assign_adjacency_aware currently only runs once, at seeding.\n\nProposal: a large-neighbourhood-search move that periodically un-divides a whole wing/subtree of the CURRENT BEST individual and reconstructs just that region using the same proven adjacency-aware constructive heuristic (seeded from the surviving circulation spine as fixed_circ, same mechanism lift_base_to_storeys already uses for upper floors), instead of relying only on small local mutation operators to find improvements. Reuses the one technique with a real track record, applied repeatedly during search rather than once at initialisation.\n\nA/B against the current baseline on programme-house and harbor-house at a fixed worker count (see 12.4's determinism-fix note about serial vs parallel admission order before trusting sub-±3 effects).","notes":"Implementation landed (uncommitted, pending A/B): operators.mutate_ruin_recreate — picks a divided, live-cut wing of one storey (2..half its leaves), un-divides it, regrows+retypes it via a scope-generalised operators._assign_adjacency_aware (new 'scope' param restricts retyping to a leaf subset while fixed_circ seeds can be border leaves outside that subset), seeded from already-typed circulation leaves bordering the wing. Room-code budget inside the wing is preserved exactly; circ/outside counts rebuilt at circ_divisor=3/outside_divisor=3. Gated like reassociate/bridge_circulation: mutation_weights['ruin_recreate']=0.0 unless driver.search(enable_ruin_recreate=True); CLI flag --ruin-recreate/--no-ruin-recreate added to evolve.py (default off). 297 existing tests pass unchanged; added smoke coverage (200 applications on harbor-house constructive seeds: zero missing-space regressions, all canonical). A/B now running in background: experiments/run_f1d_ab.sh, qpk protocol (harbor-house budget=2500 seeds 1-3, programme-house budget=3000 seeds 1-5, workers=4, both arms finished with standard --collapse), results -\u003e scratch/f1d_ab_results.tsv.","status":"closed","priority":2,"issue_type":"task","assignee":"Bruno Postle","owner":"bruno@postle.net","created_at":"2026-07-25T20:18:02Z","created_by":"Bruno Postle","updated_at":"2026-07-26T08:30:13Z","started_at":"2026-07-25T20:20:56Z","closed_at":"2026-07-26T08:30:13Z","close_reason":"DONE (positive, size-dependent). operators.mutate_ruin_recreate landed: un-divides one wing of a storey (2..half its leaves), rebuilds it via a scope-generalised _assign_adjacency_aware seeded from bordering circulation, mirroring lift_base_to_storeys' core-inheritance mechanism. Gated like reassociate/bridge_circulation (enable_ruin_recreate, default off; --ruin-recreate CLI flag). Initial uniform-weight A/B was null but underpowered (op fired ~1/32 children); a weight=3.0 follow-up (_MUTATION_WEIGHTS['ruin_recreate']=3.0, kept in source) showed a statistically significant win on programme-house across 15 seeds (8W/1L/6T, mean fails 7.07-\u003e6.00, Wilcoxon p=0.041, sign-test p=0.020) but no consistent effect on harbor-house across 8 seeds (3W/2L/3T, mean fails 73.0-\u003e74.5, slight negative lean). Kept default OFF pending a size-threshold follow-up (not filed) -- same conservative bar qpk/1ph applied before its own larger-N confirmation. Full writeup: DESIGN.md §23.","dependency_count":0,"dependent_count":0,"comment_count":0} {"id":"homemaker-py-mi7","title":"Prototype: 3D bubble-diagram relaxation of programme adjacency as a fitness signal","description":"Explore building a spring/force relaxation over the programme's required-space adjacency graph (multiple random-restart solutions), then score how well an actual Dom layout's real adjacency-graph distances correlate with a relaxed target's distances. Goal: an additional fitness term / search-guidance signal beyond the existing binary adjacency checks in graph.py. Prototype module: bubble.py. Validate by correlating similarity score against existing fitness .score on examples/programme-house's 36 scored .dom files.","notes":"Harbor-house real trajectory (driver.search, budget=6000, n=75 recorded individuals, fitness 3e-28 -\u003e 3.9e-17, fails 83-\u003e51):\n- embedding similarity(): spearman=0.164 p=0.16 (n.s.)\n- topological_similarity(): spearman=-0.160 p=0.17 (n.s.)\n\nFINAL PICTURE across 4 tests (2 programmes x 2 metric formulations, plus 2 canned-batch tests earlier): no statistically significant correlation anywhere between either the spring/embedding bubble-diagram similarity or the pure topological/abstract-graph-fitting similarity, and existing programme-driven fitness score. programme-house's real-trajectory result (rho=0.05 / rho=-0.06, n=100) is the cleanest data point — that programme has zero multi-count anonymous codes, so matched_leaves' known centroid-order matching heuristic cannot be confounding it, and it's still flat. Harbor-house is noisier (heavy anonymous counts: n x5, m x3, t x6, r x10, of x2 — the fixed centroid-order matching there is a real, uncontrolled confound) but tells the same story.\n\nRecommendation: do not pursue graph-relaxation-derived or pure-topological adjacency-matching as a fitness signal for this project without a fundamentally different formulation — two independent formulations, tested on two programmes with real evolved trajectories (not just static examples), both came back null. If revisited later, the harbor-house confound (anonymous-code instance matching) would need a real assignment solver (Hungarian/brute-force per homemaker-py-9o5's CLASS_CAP pattern) before drawing any conclusion there specifically, but programme-house's clean null already argues against the core idea. bubble.py is left in the repo (uncommitted) as a documented, working prototype/reference — not wired into fitness.py.","status":"closed","priority":2,"issue_type":"task","assignee":"Bruno Postle","owner":"bruno@postle.net","created_at":"2026-07-25T12:42:02Z","created_by":"Bruno Postle","updated_at":"2026-07-25T20:12:20Z","started_at":"2026-07-25T12:42:26Z","closed_at":"2026-07-25T20:12:20Z","close_reason":"Two independent formulations (spring/embedding bubble-diagram, pure topological hop-distance) both tested null against real evolve trajectories on programme-house (n=100, clean — no anonymous-code confound) and harbor-house (n=75). No positive correlation with existing fitness score found. bubble.py left in repo, uncommitted, as documented reference; not wired into fitness.py. Consistent with the project's broader pattern (see DESIGN.md §11.4/11.5/12.3/12.4/14/16/21/22): 'search machinery' / fitness-shaping changes have been null-to-negative essentially every time they've been tried; only construction/seeding quality and representation-relaxation changes (leaf-sharing, type superposition, global collapse) have ever moved the needle. This session's result is another data point for that pattern, not an exception.","dependency_count":0,"dependent_count":0,"comment_count":0} {"id":"homemaker-py-qi6","title":"Circulation placement to clear not-connected / access fails","description":"The 94g finish-time collapse cannot touch the \"not-connected\" (level N not\nconnected) and related access/inaccessible fails: they are properties of the\nCIRCULATION skeleton (c/o/s cells), which the collapse deliberately never\nrelabels (they form the structure the room assignment is layered onto). On the\nharbor-house best layout 2 of the 15 fails are not-connected (levels 0 and 1);\nthese are out of scope for any label optimisation.\n\nGOAL: a search/repair step that places or reshapes circulation so every usable\nspace is reachable and each storey's circulation graph is connected. Candidate\nmechanisms: (a) a mutation/operator that inserts a circulation cell to bridge a\ndisconnected component (graph.py already computes connected components +\nconnected_circulation); (b) a finish-time repair that re-types a boundary cell to\ncirculation where it reconnects the graph at least net-fail cost; (c) bias the\nouter search toward connected topologies via the graded signal.\n\nInteracts with 94g: circulation placement changes which cells are skeleton vs\nassignable, so it should run BEFORE the label collapse (collapse then optimises\nlabels over the improved skeleton). Also interacts with the public-access pin\n(94g) — better circulation placement can supply invariant inside public access,\nremoving the need to pin a room provider.\n\nMeasure on the 6 evolved layouts from the 94g sweep (not-connected + access +\ninaccessible fail counts). Related: 94g, homemaker-py-2g5.","notes":"LANDED (2026-07-18) mechanism (c) — graded circulation-connectivity signal (DESIGN §18). graph.circulation_connectivity(G) = largest-circ-component fraction [0,1]; summed over storeys it rides the score_with_grade proximity channel, gated by conf flag conn_grade (replaces the §11.4 leaf-grade on that channel). Secondary comparator key (-n_fails, grade, fitness) only — scalar fitness and fail count byte-identical (verified). Wired conn_grade through driver _overrides_for/_fitness_for/_evaluate/search (enabling it implies the grade key); evolve --conn-grade (HOMEMAKER_CONN_GRADE, default OFF). 9 new tests (tests/test_conn_grade.py), 276 pass; 60-eval CLI smoke confirms plumbing.\n\nA/B VERDICT (2026-07-22, qpk protocol, experiments/run_qi6_ab.sh) — NEGATIVE. conn_grade ON vs OFF, full-budget, both finished with --collapse: harbor-house (budget 2500, seeds 1-3) byte-identical output in every seed — the grade never fired. programme-house (budget 3000, seeds 1-5) 3/5 seeds tie exactly; seeds 1/2 diverge to a different topology with one fewer total fail, but the diff is adjacency/crinkliness/width/access/size, not connectivity. Zero cases (of 4) where a not-connected fail was present and cleared by the grade. Mechanism (b)/(c) (graded proximity as tertiary comparator key) is falsified, not just unconfirmed. Kept default OFF (already was). DESIGN.md §18 updated with full verdict.\n\nRemaining candidate: mechanism (a), an explicit insert/relocate-circulation mutation/repair operator that doesn't depend on the search stumbling onto a fail-count tie. Not started — filing as follow-on if this gets picked up; otherwise low priority (fitness fidelity, not search capability, per 94g framing).","status":"closed","priority":2,"issue_type":"feature","assignee":"Bruno Postle","owner":"bruno@postle.net","created_at":"2026-07-18T10:12:27Z","created_by":"Bruno Postle","updated_at":"2026-07-23T17:22:14Z","started_at":"2026-07-18T12:17:08Z","closed_at":"2026-07-23T17:22:14Z","close_reason":"A/B measured negative (see notes + DESIGN.md §18); mechanism (a) follow-on filed as homemaker-py-8sh","dependencies":[{"issue_id":"homemaker-py-qi6","depends_on_id":"homemaker-py-94g","type":"discovered-from","created_at":"2026-07-18T11:12:27Z","created_by":"Bruno Postle","metadata":"{}"}],"dependency_count":0,"dependent_count":0,"comment_count":0} @@ -93,26 +93,27 @@ {"id":"homemaker-py-erc.6","title":"Experiment: inner-loop slack-expansion objective term","description":"Inner-loop counterpart to plot-fill construction. If Diagnostic B shows the inner loop has room to expand leaves into slack but no objective gradient to do so (the scalar rewards hitting target area but not exceeding it where slack exists), add a term/incentive so the ratio optimiser pushes leaf boundaries out to consume neighbouring slack and satisfy size, rather than parking at target.\n\nCONDITIONAL on Diagnostic B: build this only if B localizes the gap to the inner loop (room to expand, no gradient); if B shows construction targets too-small dims, prefer the plot-fill construction sibling. Must preserve the §5.4 inner-loop cliff / §4.9 lexicographic protection — the term sits where it cannot displace the fail-count ordering. A/B vs §12.2 baseline, seeds 0/1/2, 20000 evals, staged, default-OFF. Record DESIGN.md §13.6.","notes":"DEPRIORITISED by Diagnostic B (§13.2). B shows the inner loop CANNOT repair undersize: the slack is depth-driven maldistribution baked into the frozen topology, and the equal-offset ratio DOF cannot shrink a 14x leaf to feed a starved one without trading into shape fails (0.5^n cliff). Wrong DOF and wrong direction — the blocker is slicing POSITION, not a missing expansion reward. Fix belongs upstream in construction/topology (erc.4 re-scoped, erc.3). Keep as a low-priority follow-up only if a depth-balanced construction still leaves a residual size gradient the inner loop could pick up.","status":"closed","priority":4,"issue_type":"feature","owner":"bruno@postle.net","created_at":"2026-06-22T23:16:24Z","created_by":"Bruno Postle","updated_at":"2026-06-28T13:22:22Z","closed_at":"2026-06-28T13:22:22Z","close_reason":"wont-fix (DESIGN §13.7): Diag B (§13.2) showed the inner loop cannot repair undersize (wrong DOF — slicing position, frozen-topology ratios). Superseded by depth-balanced construction (erc.4). Condition unmet.","dependencies":[{"issue_id":"homemaker-py-erc.6","depends_on_id":"homemaker-py-erc","type":"parent-child","created_at":"2026-06-23T00:16:23Z","created_by":"Bruno Postle","metadata":"{}"},{"issue_id":"homemaker-py-erc.6","depends_on_id":"homemaker-py-erc.2","type":"blocks","created_at":"2026-06-23T00:16:47Z","created_by":"Bruno Postle","metadata":"{}"}],"dependency_count":1,"dependent_count":0,"comment_count":0} {"id":"homemaker-py-erc.5","title":"Experiment: compactness-aware cuts (minimize leaf perimeter/area)","description":"Attacks the #1 factor, crinkliness (346) — a per-leaf perimeter/area property DISTINCT from proportion (aspect ratio). Proportion-aware seeding (leu.2) sizes splits but does not bias toward balanced, square-ish subdivision. Add a KD-tree-style 'keep both children compact' cut rule (prefer the cut orientation/position that minimises summed child perimeter/area) in construction.\n\nCONDITIONAL on Diagnostic A: if A shows per-leaf shape-fail is FLAT across densities (floor intrinsic to slicing density), better cuts at the same leaf count will not pay → this should be closed wont-fix in favour of leaf-sharing. Only build if A shows shape-fail RISES with density. A/B vs §12.2 baseline, seeds 0/1/2, 20000 evals, staged, default-OFF. Record DESIGN.md §13.5.","notes":"DEPRIORITISED by erc.1 verdict (§13.1): per-leaf shape-fail flat vs slicing density and cuts already squarest (_size_divisions_from_targets picks squarest rotation) yet still ~1.8 fails/leaf =\u003e little compactness headroom at fixed leaf count. Floor is intrinsic to leaf COUNT, not cut quality. Revisit only if leaf-sharing (erc.3) underdelivers.","status":"closed","priority":4,"issue_type":"feature","owner":"bruno@postle.net","created_at":"2026-06-22T23:16:21Z","created_by":"Bruno Postle","updated_at":"2026-06-28T13:22:17Z","closed_at":"2026-06-28T13:22:17Z","close_reason":"wont-fix (DESIGN §13.7): Diag A (§13.1) showed the floor is intrinsic to leaf COUNT not cut quality; revisit condition was 'only if leaf-sharing underdelivers' but leaf-sharing OVER-delivered (−32…−39%, §13.3). Condition unmet.","dependencies":[{"issue_id":"homemaker-py-erc.5","depends_on_id":"homemaker-py-erc","type":"parent-child","created_at":"2026-06-23T00:16:21Z","created_by":"Bruno Postle","metadata":"{}"},{"issue_id":"homemaker-py-erc.5","depends_on_id":"homemaker-py-erc.1","type":"blocks","created_at":"2026-06-23T00:16:43Z","created_by":"Bruno Postle","metadata":"{}"}],"dependency_count":1,"dependent_count":0,"comment_count":0} {"id":"homemaker-py-2g5","title":"Rebuild occlusion/daylight/sun subsystem in Python (post-Phase-5, after optimisation fully native)","description":"DESIGN.md §6 port scope — a whole subsystem, not a term. quality_daylight (Leaf.pm:281-296) needs Urb::Misc::Sun + Urb::Field::Occlusion (+CIESky); quality_uncrinkliness also takes the occlusion object. Indoor spaces return 1 for daylight; cost is outdoor spaces + crinkliness. Port Sun_horizontal (262980-minute normalisation) and the occlusion wall set from Dom-\u003eWalls.","acceptance_criteria":"Daylight and crinkliness factors match Perl (float tolerance) across the corpus, including multi-storey cases","notes":"Re-scoped 2026-06-12: occlusion disabled in the Urb oracle instead of ported (see homemaker-py-gp2). Native fitness ships with simple crinkliness (illumination factor = 1, in homemaker-py-gnw). This issue is now the eventual Python occlusion rebuild, only after optimisation works entirely in Python. Restores outdoor-daylight and shaded-wall selection pressure.\nReframed 2026-06-17: orthogonal to epic homemaker-py-c4c. This is fitness FIDELITY (restoring daylight + shaded-wall selection pressure to match Perl), not search CAPABILITY — it changes what 'good' means, not the search's ability to find good. It will NOT improve final designs in the sense currently sought. Stays P4, deferred until the topology-search-quality epic lands and optimisation is fully native.","status":"open","priority":4,"issue_type":"feature","owner":"bruno@postle.net","created_at":"2026-06-11T23:38:25Z","created_by":"Bruno Postle","updated_at":"2026-06-17T19:14:48Z","dependency_count":0,"dependent_count":0,"comment_count":0} -{"_type":"memory","key":"cli-tool-style-prefer-python-m-homemaker-module","value":"CLI tool style: prefer python -m homemaker.module --parameters pattern, installable via pip install -e . with pyproject.toml entry_points. Not standalone bin/ scripts."} -{"_type":"memory","key":"never-use-corpus-filenames-candidate-001-dom-candidate","value":"Never use corpus filenames (candidate-001.dom, candidate-002.dom, generated.dom, init.dom, etc.) as --output targets when running experiments. These are test fixtures. Always write experimental outputs to scratch/ or a timestamped path. Lesson from 2026-06-14: warm-start runs overwrote candidate-001/002.dom and broke graph tests."} -{"_type":"memory","key":"programme-house-optimisation-result-2026-06-14-15","value":"Programme-house optimisation result (2026-06-14/15): best achievable is 1 fail (l1 wrong level, score ~0.005). 0 fails is geometrically impossible: l1 (min 27m²) must occupy ll (~23m²) at level 0, which eliminates the t3-adj-C provider; dividing ll into lll(l1)+llr(C) gives llr proportion ~6:1 (fails). Python memetic optimizer achieves 1 fail in 50k evals vs Perl optimiser's 2-3 fails. Winning topology: TWO C nodes at level 0 — ll(C) for t3-adj-C via geometric contact, rl(C) for staircase via tree-sibling adjacency to rrr(O). Best .dom: scratch/from-warmstart-fixed.dom and scratch/from-compound3-fixed.dom."} -{"_type":"memory","key":"unfold-strategy-for-shared-leaves-homemaker-py-8iv","value":"Unfold strategy for shared leaves (homemaker-py-8iv, resolved 2026-07-16): use the BALANCED GRID (operators._grow_balanced/_size_subtree_equal), NOT circulation-aware slicing. Slicing a shared leaf perpendicular to its access edge so every child touches the corridor was implemented + A/B-tested and LOST decisively (150k-eval warm-start polish from evolved-3M: slice 41 fails/3.5e-14 vs grid 25 fails/2.4e-09, grid ahead at every milestone). Reason: k rooms all touching one wall are intrinsically thin slices; that geometric debt (proportion/long/width) is unfixable without topology change, while the grid's squarer children let local search re-route access cheaply via level_retype/place_missing/level_fix. Lesson: at the sharing-\u003eno-sharing transition, prioritise squarer children and leave access to local search; do not reintroduce slicing in Schedule B (kpu)."} {"_type":"memory","key":"adjacency-in-binary-slicing-tree-is-structural-not","value":"Adjacency in binary slicing tree is structural, not geometric: the inner-loop NM cannot fix topological adjacency failures. Two paths exist: (1) tree-sibling adjacency — a node is adjacent to its sibling in the tree; (2) cross-zone geometric adjacency — leaves from different subtrees that happen to share a boundary. Staircase/adjacency fails require a topology mutation that changes which nodes are siblings or which zones touch. This was proved empirically on programme-house: staircase fail from rot=0 layout could not be fixed by NM but was fixed by level_retype creating a two-C topology (2026-06-14/15)."} -{"_type":"memory","key":"correction-to-urb-fitness-bug-memory-bruno-2026","value":"CORRECTION to urb-fitness-bug memory (Bruno, 2026-06-12): 'C' is NOT a 'covered' type — Is_Covered is a geometric predicate (indoor space above). Urb's generic types are canonically UPPERCASE: C=circulation, O=outside, S=sahn (get_space_types qw/C O S/; corpus is 100% uppercase, never 'c'/'o' leaves). The mixed-case designs that fired the latent ratio_type first-match bug were created by homemaker's own operator type pool emitting lowercase 'c'/'o' — fixed: driver/operators now emit uppercase generics only, and class checks use t[0].lower() in 'cos'. The Urb class-sum patch stays as defensive hardening (zero impact on canonical designs). Native port (3y7/gnw): treat type classes case-insensitively, generics canonically uppercase."} -{"_type":"memory","key":"run-to-run-reproducibility-in-homemaker-layout-serial","value":"Run-to-run reproducibility in homemaker-layout: serial search (workers=1) is byte-for-byte deterministic; parallel (workers\u003e1) is now deterministic too AFTER fixing driver._run_batch to admit futures in submission order (was as_completed/completion order, bug xcy). Reproducibility holds only for a FIXED worker count — serial vs parallel differ because children-per-iteration is 1 vs n_workers (different batch granularity), which is expected, not a bug. The constructive seeder was NEVER nondeterministic: _assign_adjacency_aware has unique idx tiebreaks; comparing topologies with Python builtin hash() of the signature STRING is invalid (PYTHONHASHSEED salts str hashing per process) — use a stable hash (sha1) or genome.signature equality."} -{"_type":"memory","key":"urb-fitness-bug-found-fixed-2026-06-12","value":"Urb fitness bug found+fixed 2026-06-12 (patch in /home/bruno/src/urb, uncommitted): ProgrammeDriven.pm ratio_o/ratio_type grepped case-insensitively over the ratios hash and took the FIRST key — nondeterministic (x4.5 score swings) for designs with mixed-case type classes (both 'c' circulation and 'C' covered). Fixed to SUM the class (matches Is_Circulation//Is_Outside semantics); 35/35 corpus scores unchanged. CRITICAL for homemaker-py-3y7/gnw: the native port must implement class-SUM ratios. Building.pm has the same unpatched pattern (site-driven path, not used by our oracle). Also: the memetic search reward-hacked this bug before the fix — search results predating it are noise artifacts."} -{"_type":"memory","key":"collapse-global-94g-and-any-label-usage-optimisation","value":"collapse_global (94g) and any label/usage optimisation CANNOT fix geometry-intrinsic fails. The harbor-house 15-fail best layout contains long-thin cells that are useless whatever room usage is assigned — their width/proportion/crinkliness fails are shape-bound, not label slack. Two consequences: (1) do not over-claim collapse gains — only ~2-3 of that layout's fails are reclaimable relabel slack, the rest are geometry- or building-level bound; (2) the threshold objective must not be tuned to 'pass' a degenerate cell via a permissive room type — a metric-pass on a physically useless space is gaming, not a fix. Real remedies for these are geometry/topology search (cell shape) and circulation placement, filed separately, not the collapse."} -{"_type":"memory","key":"9o5-multi-use-leaves-is-path-a-superposition","value":"9o5 multi-use leaves is path (a) — superposition as SEARCH RELAXATION that COLLAPSES to specific usage at the end, NOT path (b) loose-fit/no-collapse. Bruno's intent: codes with SIMILAR leaf requirements form an interchangeable equivalence class; during evolution the solver doesn't commit which leaf serves which specific usage (smoother landscape, no fighting over exact leaf usage); at the end the layout is CONDENSED to specific usages by brute-forcing the in-class assignment (3 interchangeable usages over 3 leaves = 3! = 6 combinations to check, pick best). 'Derive automatically' compatibility = requirement-similarity grouping. This reverses the issue's stated 'path b preferred' note."} -{"_type":"memory","key":"user-preference-bruno-this-is-a-fedora-system","value":"User preference (Bruno): this is a Fedora system — NEVER install Python packages via pip without asking first; always ask whether to install the rpm via dnf (e.g. python3-cma) before considering pip. Applies to any dependency additions."} -{"_type":"memory","key":"warm-x0-initialization-bug-pattern-when-a-topology","value":"warm_x0 initialization bug pattern: when a topology operator explicitly sets division ratios on a newly-created node (e.g. compound_fix sets node.division=[0.25,0.25] for t3), parent.ratios has no entry for that node (it was a leaf). warm_x0 defaults it to 0.5, corrupting the inner loop's starting point and making the operator invisible to lex comparison. Fix: only propagate child ratios for nodes where the parent node was NOT already divided; stale hidden nodes revealed by structural mutations (swap flipping b.below) must NOT contribute their pre-writeback values. See driver.py lines 259-267 (fixed 2026-06-14)."} -{"_type":"memory","key":"deceptive-valleys-in-topology-search-when-every-single","value":"Deceptive valleys in topology search: when every single-step mutation from a target state passes through a high-fail intermediary (e.g. level_fix displaces a room into 5+ new fails), a compound operator that atomically applies two coordinated changes can escape. Design compound operators to land on the low-fail state directly, bypassing the deceptive gradient. Programme-house example: level_compound_fix atomically moves the level-constrained room AND re-inserts the displaced room adjacent to C in one step (operators.py, 2026-06-14)."} -{"_type":"memory","key":"multi-storey-staircase-consistency-when-dividing-or-retyping","value":"Multi-storey staircase consistency: when dividing or retyping a circulation (C) leaf at one level, the same structural change should be propagated to the matching leaf on ALL other storeys so the stair core path is maintained. The optimizer cannot fix staircase disruptions through trial-and-error geometry alone — it requires a synchronized multi-level operator that applies the same topology change to every storey simultaneously."} -{"_type":"memory","key":"urb-oracle-nondeterminism-urb-fitness-pl-output-varies","value":"Urb oracle nondeterminism: urb-fitness.pl output varies run-to-run from Perl hash-order randomisation — .fails line ORDER shuffles (compare sorted, use oracle.Score.fail_lines) and the score float can flip by ~1 ULP (compare with math.isclose rel_tol=1e-12, never ==). Not a batching artifact; affects single runs too. Matters for the Phase 3 native-fitness parity gate (homemaker-py-uxz)."} +{"_type":"memory","key":"cli-tool-style-prefer-python-m-homemaker-module","value":"CLI tool style: prefer python -m homemaker.module --parameters pattern, installable via pip install -e . with pyproject.toml entry_points. Not standalone bin/ scripts."} {"_type":"memory","key":"experiment-harness-gotcha-the-leaf-sharing-relaxed-objective","value":"Experiment harness gotcha: the leaf-sharing RELAXED objective (§13.3) is injected ONLY by monkeypatching fitness.load_config in the parent process (run_staged_search.py / probe scripts). This is parent-process-only and does NOT propagate into ProcessPoolExecutor workers (n_workers\u003e1), which re-import fitness fresh and score under the STRICT on-disk patterns.config -\u003e r.n_fails MISMATCH (worker strict vs parent relaxed re-score). ALL §13.x floor runs were therefore SERIAL. Any future PARALLEL leaf-sharing experiment will silently mis-score until leaf_sharing lives on disk/CLI (tracked: homemaker-py-x3b). The parallel driver itself is correct; both paths score via load_config(programme_dir)."} {"_type":"memory","key":"experiment-seeding-pitfall-run-search-scaled-py-s","value":"Experiment seeding pitfall: run_search_scaled.py's default PH_SEED (c964…dom) is a FINISHED programme-house design — passing it warm-starts and floors at ~3 fails, NOT a blank-slate topology search. For blank-slate runs comparable to §11.5/§11.6 baselines, seed from examples/programme-house/init.dom (a bare undivided plot; driver bootstrap auto-triggers only on bare plots). Bit the 6zy sweep — first pass used c964 and falsely showed 3-fail floor across the whole grid."} {"_type":"memory","key":"ld2-13-6-interior-o-seed-diagnostic-all","value":"ld2/§13.6 interior-O seed diagnostic: ALL crinkliness fails in the constructed bal+share seed are UNDER-exposed (crink\u003c0.62, landlocked rooms with no facade + no uncovered-O neighbour) — zero over-exposed sliver fails. So the erc crinkliness residual is genuine under-daylighting, validating the interior light-well premise. Default outside_divisor=6 was too sparse (null: harbor 147-\u003e142, crinkliness even rose). odiv=3 is the seed-optimal joint setting: harbor seed fails 147-\u003e129 (-18), maple 219-\u003e206 (-14), landlocked fails drop, at cost of more leaves (harbor +4, maple +8). Because it ADDS leaves it carries the §13.4 wash-out risk; A/B to convergence pending."} -{"_type":"memory","key":"strategy-decision-2026-06-12-bruno-occlusion-daylight","value":"Strategy decision 2026-06-12 (Bruno): occlusion/daylight is ORTHOGONAL to building a scalable optimiser. Disable it in Urb (env flag, homemaker-py-gp2) rather than port it; native fitness uses simple crinkliness (illumination factor = 1); rebuild occlusion in Python only after optimisation is fully native (homemaker-py-2g5, now P4). Consequence: all scores change when the flag flips — re-baseline corpus/.score, DESIGN \\$4.5 gains, gate bars at one clean boundary AFTER homemaker-py-1p0 closes; Phase-2 urb-evolve benchmark must run with the same flag."} -{"_type":"memory","key":"homemaker-py-3l6-fix-leaf-sharing-evolve-runs","value":"homemaker-py-3l6 fix: leaf-sharing evolve runs now auto-finish before write via driver.polish_finish — unfold_shared_leaves() then a warm-started leaf_sharing=False polish search (--polish-budget, default budget//2). Makes the written .dom honest under canonical homemaker-fitness (internal==canonical when leaf_sharing off). Interrupt path forces polish_budget=0 (unfold+rescore only). This is yaa's unfold-then-polish, made automatic; Schedule B annealing is still kpu."} +{"_type":"memory","key":"programme-house-optimisation-result-2026-06-14-15","value":"Programme-house optimisation result (2026-06-14/15): best achievable is 1 fail (l1 wrong level, score ~0.005). 0 fails is geometrically impossible: l1 (min 27m²) must occupy ll (~23m²) at level 0, which eliminates the t3-adj-C provider; dividing ll into lll(l1)+llr(C) gives llr proportion ~6:1 (fails). Python memetic optimizer achieves 1 fail in 50k evals vs Perl optimiser's 2-3 fails. Winning topology: TWO C nodes at level 0 — ll(C) for t3-adj-C via geometric contact, rl(C) for staircase via tree-sibling adjacency to rrr(O). Best .dom: scratch/from-warmstart-fixed.dom and scratch/from-compound3-fixed.dom."} +{"_type":"memory","key":"collapse-global-s-jacobi-adjacency-relaxation-homemaker-py","value":"collapse_global's Jacobi adjacency relaxation (homemaker-py-94g) is a synchronous per-round linear-assignment re-solve, which can 2-cycle indefinitely between two labellings that each satisfy ZERO adjacency requirements even though a permutation satisfying ALL of them exists -- proven on a minimal 4-cell chain (p1-q1-p2-q2, two disjoint adjacency pairs p1\u003c-\u003ep2/q1\u003c-\u003eq2) in test_two_opt_polish_escapes_jacobi_plateau. homemaker-py-9wi added Fitness._two_opt_adjacency_polish: a same-level pairwise-swap local search run after the Jacobi fixpoint, gated behind collapse_global(local_search=True) (default off, exposed as homemaker-collapse --local-search). Monotone by construction (a swap is kept only if it strictly increases total reward). Empirically on the 11 harbor-house evolved-*.dom/3m.dom/materialised-3M.dom layouts: 10 matched Jacobi-only exactly, 0 regressed, and evolved-anneal-3M.dom improved 21-\u003e19 fails (fixed a genuine mutual da1\u003c-\u003ek1 adjacency miss the Jacobi loop couldn't reach)."} +{"_type":"memory","key":"urb-fitness-bug-found-fixed-2026-06-12","value":"Urb fitness bug found+fixed 2026-06-12 (patch in /home/bruno/src/urb, uncommitted): ProgrammeDriven.pm ratio_o/ratio_type grepped case-insensitively over the ratios hash and took the FIRST key — nondeterministic (x4.5 score swings) for designs with mixed-case type classes (both 'c' circulation and 'C' covered). Fixed to SUM the class (matches Is_Circulation//Is_Outside semantics); 35/35 corpus scores unchanged. CRITICAL for homemaker-py-3y7/gnw: the native port must implement class-SUM ratios. Building.pm has the same unpatched pattern (site-driven path, not used by our oracle). Also: the memetic search reward-hacked this bug before the fix — search results predating it are noise artifacts."} +{"_type":"memory","key":"urb-oracle-nondeterminism-urb-fitness-pl-output-varies","value":"Urb oracle nondeterminism: urb-fitness.pl output varies run-to-run from Perl hash-order randomisation — .fails line ORDER shuffles (compare sorted, use oracle.Score.fail_lines) and the score float can flip by ~1 ULP (compare with math.isclose rel_tol=1e-12, never ==). Not a batching artifact; affects single runs too. Matters for the Phase 3 native-fitness parity gate (homemaker-py-uxz)."} +{"_type":"memory","key":"user-preference-bruno-this-is-a-fedora-system","value":"User preference (Bruno): this is a Fedora system — NEVER install Python packages via pip without asking first; always ask whether to install the rpm via dnf (e.g. python3-cma) before considering pip. Applies to any dependency additions."} {"_type":"memory","key":"homemaker-py-pythonpath-set-pythonpath-home-bruno-src","value":"homemaker-layout PYTHONPATH: package installed as 'homemaker-layout' via pip install -e . so 'import homemaker_layout' works from anywhere without PYTHONPATH. For running tests use 'python -m pytest' from project root /home/bruno/src/homemaker-layout (pyproject.toml adds src/ automatically). Never try pip show homemaker — that's the old homemaker-addon conflict."} -{"_type":"memory","key":"island-model-psk-14-is-a-null-priming","value":"Island model (psk, §14) is a NULL: priming a population from N converged independent elites + crossover-heavy migration does not beat best-of-N at equal total budget (maple island 124 vs control 116). The child_probe instrument shows WHY: area-matched crossover across independently-converged elites almost never synthesizes (1-3 of ~64 children beat the better parent, max drop 2-5) because the slicing encoding is non-canonical (9gp), so splices are disruptive not combinatorial. Search-machinery null #3 after graded-objective and niching/restarts; residual stays geometry/shape-bound."} {"_type":"memory","key":"proportion-aware-constructive-seeding-leu-2-12-2","value":"Proportion-aware constructive seeding (leu.2/§12.2): sizing seed cuts from target AREAS only regresses (thin slivers wreck aspect); you must ALSO pick each cut's rotation for child squareness. It is a convergence ACCELERATOR via a deeper local optimum around the constructed topology: wins where that topology is roughly right and budget is scarce (harbor -13%, maple -10% at 20k evals) but DELAYS small programmes where the seed must be restructured by undivide (programme-house regresses at fixed budget, yet reaches the floor given budget - speed, not asymptote). Default-on. Also: n_storeys must honour storey_minimum, not just level: keys (programme-house storey_minimum:2, all rooms level:0 - was seeded 1 storey short; cq1)."} +{"_type":"memory","key":"unfold-strategy-for-shared-leaves-homemaker-py-8iv","value":"Unfold strategy for shared leaves (homemaker-py-8iv, resolved 2026-07-16): use the BALANCED GRID (operators._grow_balanced/_size_subtree_equal), NOT circulation-aware slicing. Slicing a shared leaf perpendicular to its access edge so every child touches the corridor was implemented + A/B-tested and LOST decisively (150k-eval warm-start polish from evolved-3M: slice 41 fails/3.5e-14 vs grid 25 fails/2.4e-09, grid ahead at every milestone). Reason: k rooms all touching one wall are intrinsically thin slices; that geometric debt (proportion/long/width) is unfixable without topology change, while the grid's squarer children let local search re-route access cheaply via level_retype/place_missing/level_fix. Lesson: at the sharing-\u003eno-sharing transition, prioritise squarer children and leave access to local search; do not reintroduce slicing in Schedule B (kpu)."} +{"_type":"memory","key":"never-use-corpus-filenames-candidate-001-dom-candidate","value":"Never use corpus filenames (candidate-001.dom, candidate-002.dom, generated.dom, init.dom, etc.) as --output targets when running experiments. These are test fixtures. Always write experimental outputs to scratch/ or a timestamped path. Lesson from 2026-06-14: warm-start runs overwrote candidate-001/002.dom and broke graph tests."} +{"_type":"memory","key":"run-to-run-reproducibility-in-homemaker-layout-serial","value":"Run-to-run reproducibility in homemaker-layout: serial search (workers=1) is byte-for-byte deterministic; parallel (workers\u003e1) is now deterministic too AFTER fixing driver._run_batch to admit futures in submission order (was as_completed/completion order, bug xcy). Reproducibility holds only for a FIXED worker count — serial vs parallel differ because children-per-iteration is 1 vs n_workers (different batch granularity), which is expected, not a bug. The constructive seeder was NEVER nondeterministic: _assign_adjacency_aware has unique idx tiebreaks; comparing topologies with Python builtin hash() of the signature STRING is invalid (PYTHONHASHSEED salts str hashing per process) — use a stable hash (sha1) or genome.signature equality."} +{"_type":"memory","key":"9o5-multi-use-leaves-is-path-a-superposition","value":"9o5 multi-use leaves is path (a) — superposition as SEARCH RELAXATION that COLLAPSES to specific usage at the end, NOT path (b) loose-fit/no-collapse. Bruno's intent: codes with SIMILAR leaf requirements form an interchangeable equivalence class; during evolution the solver doesn't commit which leaf serves which specific usage (smoother landscape, no fighting over exact leaf usage); at the end the layout is CONDENSED to specific usages by brute-forcing the in-class assignment (3 interchangeable usages over 3 leaves = 3! = 6 combinations to check, pick best). 'Derive automatically' compatibility = requirement-similarity grouping. This reverses the issue's stated 'path b preferred' note."} +{"_type":"memory","key":"correction-to-urb-fitness-bug-memory-bruno-2026","value":"CORRECTION to urb-fitness-bug memory (Bruno, 2026-06-12): 'C' is NOT a 'covered' type — Is_Covered is a geometric predicate (indoor space above). Urb's generic types are canonically UPPERCASE: C=circulation, O=outside, S=sahn (get_space_types qw/C O S/; corpus is 100% uppercase, never 'c'/'o' leaves). The mixed-case designs that fired the latent ratio_type first-match bug were created by homemaker's own operator type pool emitting lowercase 'c'/'o' — fixed: driver/operators now emit uppercase generics only, and class checks use t[0].lower() in 'cos'. The Urb class-sum patch stays as defensive hardening (zero impact on canonical designs). Native port (3y7/gnw): treat type classes case-insensitively, generics canonically uppercase."} +{"_type":"memory","key":"deceptive-valleys-in-topology-search-when-every-single","value":"Deceptive valleys in topology search: when every single-step mutation from a target state passes through a high-fail intermediary (e.g. level_fix displaces a room into 5+ new fails), a compound operator that atomically applies two coordinated changes can escape. Design compound operators to land on the low-fail state directly, bypassing the deceptive gradient. Programme-house example: level_compound_fix atomically moves the level-constrained room AND re-inserts the displaced room adjacent to C in one step (operators.py, 2026-06-14)."} +{"_type":"memory","key":"strategy-decision-2026-06-12-bruno-occlusion-daylight","value":"Strategy decision 2026-06-12 (Bruno): occlusion/daylight is ORTHOGONAL to building a scalable optimiser. Disable it in Urb (env flag, homemaker-py-gp2) rather than port it; native fitness uses simple crinkliness (illumination factor = 1); rebuild occlusion in Python only after optimisation is fully native (homemaker-py-2g5, now P4). Consequence: all scores change when the flag flips — re-baseline corpus/.score, DESIGN \\$4.5 gains, gate bars at one clean boundary AFTER homemaker-py-1p0 closes; Phase-2 urb-evolve benchmark must run with the same flag."} +{"_type":"memory","key":"collapse-global-94g-and-any-label-usage-optimisation","value":"collapse_global (94g) and any label/usage optimisation CANNOT fix geometry-intrinsic fails. The harbor-house 15-fail best layout contains long-thin cells that are useless whatever room usage is assigned — their width/proportion/crinkliness fails are shape-bound, not label slack. Two consequences: (1) do not over-claim collapse gains — only ~2-3 of that layout's fails are reclaimable relabel slack, the rest are geometry- or building-level bound; (2) the threshold objective must not be tuned to 'pass' a degenerate cell via a permissive room type — a metric-pass on a physically useless space is gaming, not a fix. Real remedies for these are geometry/topology search (cell shape) and circulation placement, filed separately, not the collapse."} +{"_type":"memory","key":"homemaker-py-3l6-fix-leaf-sharing-evolve-runs","value":"homemaker-py-3l6 fix: leaf-sharing evolve runs now auto-finish before write via driver.polish_finish — unfold_shared_leaves() then a warm-started leaf_sharing=False polish search (--polish-budget, default budget//2). Makes the written .dom honest under canonical homemaker-fitness (internal==canonical when leaf_sharing off). Interrupt path forces polish_budget=0 (unfold+rescore only). This is yaa's unfold-then-polish, made automatic; Schedule B annealing is still kpu."} +{"_type":"memory","key":"island-model-psk-14-is-a-null-priming","value":"Island model (psk, §14) is a NULL: priming a population from N converged independent elites + crossover-heavy migration does not beat best-of-N at equal total budget (maple island 124 vs control 116). The child_probe instrument shows WHY: area-matched crossover across independently-converged elites almost never synthesizes (1-3 of ~64 children beat the better parent, max drop 2-5) because the slicing encoding is non-canonical (9gp), so splices are disruptive not combinatorial. Search-machinery null #3 after graded-objective and niching/restarts; residual stays geometry/shape-bound."} +{"_type":"memory","key":"warm-x0-initialization-bug-pattern-when-a-topology","value":"warm_x0 initialization bug pattern: when a topology operator explicitly sets division ratios on a newly-created node (e.g. compound_fix sets node.division=[0.25,0.25] for t3), parent.ratios has no entry for that node (it was a leaf). warm_x0 defaults it to 0.5, corrupting the inner loop's starting point and making the operator invisible to lex comparison. Fix: only propagate child ratios for nodes where the parent node was NOT already divided; stale hidden nodes revealed by structural mutations (swap flipping b.below) must NOT contribute their pre-writeback values. See driver.py lines 259-267 (fixed 2026-06-14)."} +{"_type":"memory","key":"multi-storey-staircase-consistency-when-dividing-or-retyping","value":"Multi-storey staircase consistency: when dividing or retyping a circulation (C) leaf at one level, the same structural change should be propagated to the matching leaf on ALL other storeys so the stair core path is maintained. The optimizer cannot fix staircase disruptions through trial-and-error geometry alone — it requires a synchronized multi-level operator that applies the same topology change to every storey simultaneously."} diff --git a/src/homemaker_layout/collapse_cmd.py b/src/homemaker_layout/collapse_cmd.py index 507d1e0..09bf86a 100644 --- a/src/homemaker_layout/collapse_cmd.py +++ b/src/homemaker_layout/collapse_cmd.py @@ -48,6 +48,10 @@ def _parse_args(argv): action=argparse.BooleanOptionalAction, default=True, help="revert if the collapse increases the fail count " "(default: on)") + p.add_argument("--local-search", dest="local_search", + action=argparse.BooleanOptionalAction, default=False, + help="2-opt polish past the Jacobi adjacency relaxation " + "(homemaker-py-9wi, default: off)") return p.parse_args(argv) @@ -63,6 +67,7 @@ def main(argv=None) -> int: adjacency=args.adjacency, objective=args.objective, preserve_public_access=args.public_access, + local_search=args.local_search, ) rc = 0 diff --git a/src/homemaker_layout/fitness.py b/src/homemaker_layout/fitness.py index a697af7..3787aef 100644 --- a/src/homemaker_layout/fitness.py +++ b/src/homemaker_layout/fitness.py @@ -354,6 +354,124 @@ class Fitness: # quality only breaks ties; far below the forbid penalty so level holds. _COLLAPSE_FAIL_W = 1e6 + def _collapse_value( + self, + lf: Node, + code: str, + lvl: int, + prog: dict, + objective: str, + forbid: float, + fail_w: float, + ) -> float: + """Base (non-adjacency) collapse value of relabelling ``lf`` (on storey + ``lvl``) to ``code``: the ``_COLLAPSE_FORBID`` penalty on a level + mismatch, else quality_size*width*proportion*area, plus ``fail_w`` per + passing factor under the ``"threshold"`` objective. Shared by the + collapse_global assignment matrix and the 2-opt polish below so both + score a (leaf, code) pair identically.""" + req = prog[code] + if req.level is not None and req.level != lvl: + return forbid + orig = lf.type + lf.type = code + try: + qs = self.quality_size(lf) + qw = self.quality_width(lf) + qp = self.quality_proportion(lf) + finally: + lf.type = orig + val = qs * qw * qp * geometry.area(lf) + if objective == "threshold": + passes = ( + (qs >= FAIL_THRESHOLD) + (qw >= FAIL_THRESHOLD) + (qp >= FAIL_THRESHOLD) + ) + val += fail_w * passes + return val + + def _two_opt_adjacency_polish( + self, + supply: list[Node], + levels_of: list[int], + graphs: list, + code_adj: dict[str, list[str]], + prog: dict, + objective: str, + forbid: float, + fail_w: float, + max_passes: int = 20, + ) -> None: + """homemaker-py-9wi: a local-search pass beyond collapse_global's Jacobi + adjacency relaxation. Jacobi re-solves a LINEAR assignment each round + holding neighbours' labels fixed from the previous round -- exact per + round, but the true objective is quadratic (a satisfied adjacency + depends on a PAIR of labels), so synchronous Jacobi can plateau short + of the joint optimum. This adds 2-opt: for every same-level pair of + supply leaves, try swapping their CURRENT labels and keep the swap + only if it strictly increases the total reward (own quality/threshold + value + fail_w per satisfied adjacency) summed over the two leaves and + every leaf adjacent to either -- the only cells a label swap between + i and j can change. Repeats to a fixpoint (or ``max_passes``). + + Same-level-only pairing keeps the hard level constraint for free: both + codes already matched their own leaf's level before the swap, and the + two leaves share a level, so the swap is valid on both sides. A swap + is applied only when it is a STRICT improvement, so this can only + reduce, never increase, the fail count -- monotone by construction, + like the Hungarian solve it refines.""" + from . import graph as graph_mod + + idx_of_leaf = {id(lf): i for i, lf in enumerate(supply)} + + def reward(idx: int) -> float: + lf = supply[idx] + code = lf.type + val = self._collapse_value( + lf, code, levels_of[idx], prog, objective, forbid, fail_w + ) + if val <= forbid: + return val + G = graphs[levels_of[idx]] + sat = sum(1 for ac in code_adj.get(code, ()) if graph_mod.has_adjacency(lf, ac, G)) + return val + fail_w * sat + + def affected(i: int, j: int) -> set[int]: + aff = {i, j} + for k in (i, j): + lf = supply[k] + G = graphs[levels_of[k]] + if G.has_node(lf): + for nb in G.neighbors(lf): + nidx = idx_of_leaf.get(id(nb)) + if nidx is not None: + aff.add(nidx) + return aff + + by_level: dict[int, list[int]] = {} + for idx, lvl in enumerate(levels_of): + by_level.setdefault(lvl, []).append(idx) + + changed = True + passes = 0 + while changed and passes < max_passes: + changed = False + passes += 1 + for idxs in by_level.values(): + for a in range(len(idxs)): + for b in range(a + 1, len(idxs)): + i, j = idxs[a], idxs[b] + ci, cj = supply[i].type, supply[j].type + if ci == cj: + continue + aff = affected(i, j) + before = sum(reward(k) for k in aff) + supply[i].type, supply[j].type = cj, ci + after = sum(reward(k) for k in aff) + if after > before + 1e-9: + changed = True + else: + supply[i].type, supply[j].type = ci, cj + def collapse_global( self, root: Node, @@ -361,6 +479,8 @@ class Fitness: objective: str = "threshold", preserve_public_access: bool = True, iters: int = 6, + local_search: bool = False, + local_search_passes: int = 20, ) -> None: """Finish-time GLOBAL cell->room collapse (homemaker-py-94g): relabel every inside-room leaf across the whole building to the required room it @@ -397,6 +517,18 @@ class Fitness: same weight (_COLLAPSE_FAIL_W = one avoided fail), so the collapse minimises (adjacency + size/width/proportion) fails jointly. + LOCAL_SEARCH (homemaker-py-9wi, default off): after the Jacobi loop + above reaches its fixpoint, run a 2-opt polish (_two_opt_adjacency_polish) + that tries swapping the labels of every same-level pair of supply leaves + and keeps a swap only if it strictly improves the total reward. Jacobi + re-solves a LINEAR assignment each round holding neighbours' labels fixed + from the previous round, so it can plateau short of the true quadratic- + assignment optimum (a satisfied adjacency depends on a PAIR of labels, + not one); 2-opt reaches past that plateau. Monotone by construction (only + strictly-improving swaps are kept), so it is safe to try whenever + ``adjacency`` is on — enable per-run and A/B against the Jacobi-only + result before defaulting it on. + PRESERVE_PUBLIC_ACCESS pins the room leaf that solely provides the building's street access (an l/k neighbour of a public outside leaf, with no circulation fallback) so the collapse cannot drop the building-level @@ -448,34 +580,16 @@ class Fitness: forbid = self._COLLAPSE_FORBID fail_w = self._COLLAPSE_FAIL_W levels_of = [dom_mod.level_of(lf) for lf in supply] - areas = [geometry.area(lf) for lf in supply] # Base per-cell value: forbid on level mismatch, else the separable fit. # In "threshold" mode add fail_w per passing size/width/proportion factor # so the matching maximises passes first, continuous fit only as tiebreak. - base: list[list[float]] = [] - for i, lf in enumerate(supply): - orig = lf.type - row = [] - for code in slots: - req = prog[code] - if req.level is not None and req.level != levels_of[i]: - row.append(forbid) - continue - lf.type = code - qs = self.quality_size(lf) - qw = self.quality_width(lf) - qp = self.quality_proportion(lf) - val = qs * qw * qp * areas[i] - if objective == "threshold": - passes = ( - (qs >= FAIL_THRESHOLD) - + (qw >= FAIL_THRESHOLD) - + (qp >= FAIL_THRESHOLD) - ) - val += fail_w * passes - row.append(val) - lf.type = orig - base.append(row) + base: list[list[float]] = [ + [ + self._collapse_value(lf, code, levels_of[i], prog, objective, forbid, fail_w) + for code in slots + ] + for i, lf in enumerate(supply) + ] if not adjacency: for r, c in self._best_assignment(base): @@ -514,6 +628,19 @@ class Fitness: break prev_labels = new_labels + if local_search: + self._two_opt_adjacency_polish( + supply, + levels_of, + graphs, + code_adj, + prog, + objective, + forbid, + fail_w, + max_passes=local_search_passes, + ) + def _public_access_pins( self, root: Node, graphs: list, lvls: list, room_codes: set ) -> set[int]: diff --git a/tests/test_collapse_global.py b/tests/test_collapse_global.py index 483d875..a1d44d4 100644 --- a/tests/test_collapse_global.py +++ b/tests/test_collapse_global.py @@ -105,6 +105,69 @@ def test_single_code_is_noop(): assert [lf.type for lf in root.leaves()] == ["b1", "b1"] +# --------------------------------------------------------------------------- # +# 2-opt local search beyond the Jacobi plateau (homemaker-py-9wi) +# --------------------------------------------------------------------------- # + +def _four_leaf_chain(t1: str, t2: str, t3: str, t4: str, width: float = 1.0, height: float = 2.0): + # A 1x4 strip of equal cells split twice at 0.5: the leaf-adjacency graph + # is a chain (1-2, 2-3, 3-4) with no 1-3/2-4 edges -- see build_graphs. + geometry.clear_cache() + left = Node(rotation=0, division=[0.5, 0.5], left=Node(type=t1), right=Node(type=t2)) + right = Node(rotation=0, division=[0.5, 0.5], left=Node(type=t3), right=Node(type=t4)) + root = Node( + node=[[0, 0], [4 * width, 0], [4 * width, height], [0, height]], + rotation=0, division=[0.5, 0.5], + left=left, right=right, + ) + _link_subtree(root, None, "") + return root + + +def test_two_opt_polish_escapes_jacobi_plateau(): + # Two adjacency pairs (p1<->p2, q1<->q2) on a 4-cell chain p1-q1-p2-q2. + # Every code shares identical size/width/proportion targets (all four + # cells are geometrically identical), so the ONLY thing that can prefer + # one labelling over another is adjacency -- isolating the effect. + # + # Starting interleaved (p1,q1,p2,q2), the true optimum interleaves the + # OTHER way (p1,p2 adjacent + q1,q2 adjacent, 4 satisfied requirements), + # but the Jacobi relaxation (adjacency bonus computed from the PREVIOUS + # round's neighbour labels, re-solved synchronously) 2-cycles between two + # states that each satisfy 0 requirements and never reaches it -- a + # textbook case of the quadratic-assignment plateau the issue describes. + # 2-opt, tried after the Jacobi fixpoint, finds the escaping swap. + spec = { + "size": [2.0, 1.0], "width": [1.0, 1.0], "proportion": [2.0, 1.0], "count": 1, + } + conf = _conf({ + "p1": {**spec, "adjacency": ["p2"]}, + "p2": {**spec, "adjacency": ["p1"]}, + "q1": {**spec, "adjacency": ["q2"]}, + "q2": {**spec, "adjacency": ["q1"]}, + }) + fit = Fitness(conf=conf) + + def satisfied(root): + from homemaker_layout import graph as graph_mod + G = graph_mod.build_graphs(root, 1.2)[0] + prog = fit._programme + return sum( + 1 + for lf in root.leaves() + for ac in prog[lf.type].adjacency + if graph_mod.has_adjacency(lf, ac, G) + ) + + root_jacobi = _four_leaf_chain("p1", "q1", "p2", "q2") + fit.collapse_global(root_jacobi, adjacency=True, local_search=False) + assert satisfied(root_jacobi) == 0 # the Jacobi-only plateau + + root_polished = _four_leaf_chain("p1", "q1", "p2", "q2") + fit.collapse_global(root_polished, adjacency=True, local_search=True) + assert satisfied(root_polished) == 4 # 2-opt reaches the true optimum + + def test_collapse_finish_is_keep_better_and_unmerged(): # collapse_finish returns (tree, base, collapsed, applied); the tree it hands # back is unmerged (leaves still carry their divisions), and collapsed<=base.