Skip to content

feat(problems): IDTLZ1 and IDTLZ2 of Jain and Deb, as PlatEMO defines them, with reference fronts - #366

Merged
NicolasRodriguezUribe merged 2 commits into
mainfrom
feat/idtlz-problems
Nov 19, 2026
Merged

NicolasRodriguezUribe merged 2 commits into
mainfrom
feat/idtlz-problems

Conversation

@NicolasRodriguezUribe

Copy link
Copy Markdown
Contributor

Summary

Adds the inverted DTLZ problems IDTLZ1 and IDTLZ2 (idtlz1, idtlz2) as PlatEMO defines them, with two- and three-objective reference fronts (change request P3). They are the usual test of how decomposition-based algorithms (NSGA-III, MOEA/D, RVEA) cope with an inverted front, and they were missing for the planned comparison of VAMOS with jMetal, jMetalPy, pymoo and PlatEMO.

Source: H. Jain and K. Deb, "An Evolutionary Many-Objective Optimization Algorithm Using Reference-Point Based Nondominated Sorting Approach, Part II", IEEE TEVC 18(4):602–622, 2014, doi:10.1109/TEVC.2013.2281534, as implemented in PlatEMO's Problems/Multi-objective optimization/DTLZ/IDTLZ1.m and IDTLZ2.m (BIMK/PlatEMO at e2d58cb, read from the repository; no code copied).

  • Problems (src/vamos/foundation/problem/dtlz.py, IDTLZ1Problem and IDTLZ2Problem, registered in the DTLZ family as idtlz1 and idtlz2). Every objective is minimized, the variables lie in [0, 1], and n_obj is configurable as for DTLZ.
    • IDTLZ1 (CalObj of IDTLZ1.m): f = (1 + g) / 2 - 0.5 (1 + g) s, with DTLZ1's g = 100 (k + sum((x_i - 0.5)^2 - cos(20 pi (x_i - 0.5)))) over the last k = n - M + 1 variables and DTLZ1's position shape s = fliplr(cumprod([1, x_1..x_{M-1}])) .* [1, 1 - x_{M-1}..1 - x_1]. PlatEMO's D = M + 4, so n_var defaults to n_obj + 4 (7 with three objectives).
    • IDTLZ2 (IDTLZ2.m): f = (1 + g) - (1 + g) s, with g = sum (x_i - 0.5)^2 over the last variables and DTLZ2's shape (cos(x pi / 2) and sin(x pi / 2)). D = M + 9, so n_var defaults to n_obj + 9 (12).
    • The shape is computed as PlatEMO writes it (cumulative product, flipped), in one helper for both. As for DTLZ, a non-standard k warns and n_var < n_obj raises.
    • pymoo 0.6.2's InvertedDTLZ1 (0.5 (1 + g) - DTLZ1(x)) is the same function; pymoo has no inverted DTLZ2 and jMetal neither problem.
    • The classes are not exported by vamos.problems; the problems are built by name (optimize("idtlz2", ...), make_problem_selection("idtlz1", n_obj=5)), as DTLZ5 and DTLZ6 are.
  • Reference fronts idtlz1.2d.csv, idtlz1.3d.csv, idtlz2.2d.csv and idtlz2.3d.csv, written by experiments/scripts/generate_reference_fronts.py from PlatEMO's GetOptimum: (1 - w) / 2 (IDTLZ1) and 1 - w / norm(w) (IDTLZ2) for w on the unit simplex. With two objectives, 1000 points spaced evenly along the front (IDTLZ1's is DTLZ1's segment f1 + f2 = 0.5, and the file is the same as dtlz1.2d.csv); with three, the 5050-point lattice (H = 99) of the DTLZ1–DTLZ4 samples. The resolver finds them by n_obj (<key>.<M>d.csv, three objectives by default).
  • Docs: docs/reference/problems.md: the registered count (260), the class list, two catalog rows, a paragraph on the definition and the shape of the fronts, the reference-front notes and a provenance row, and the source link, all marked Development after 1.4.0:.

Not done: fronts with more objectives. They are as cheap and exact (the same formulas on a larger lattice), but the lattice convention for 4, 6 and 8 objectives arrives with the DTLZ fronts of #362; once both are merged, idtlz1.<m>d/idtlz2.<m>d can be generated with that lattice in a short follow-up, instead of a second copy of the same helper in this branch. IDTLZ3 and IDTLZ4 (also in PlatEMO) were not requested.

Tests

tests/foundation/test_idtlz_problems.py:

  • test_objectives_follow_platemos_formulas: IDTLZ1 and IDTLZ2 with 2, 3 and 5 objectives on fixed decision vectors (the bounds, the centre and five seeded points); the expected values are written out in the test from PlatEMO's formulas, one objective at a time with math (f_i uses x_1 ... x_{M-i} and the complement of x_{M-i+1}), not with the implementation's cumulative product; rtol = 1e-12.
  • test_objectives_at_the_centre_of_the_box: hand-computed values, (0.375, 0.375, 0.25) for IDTLZ1 and (0.5, 0.5, 1 - sqrt(0.5)) for IDTLZ2.
  • test_idtlz_turns_dtlz_over: IDTLZ1 = (1 + g) / 2 minus VAMOS's DTLZ1, and IDTLZ2 = (1 + g) minus DTLZ2, with 2, 3 and 6 objectives (pymoo's formulation).
  • test_reference_front_points_are_reached_with_g_zero: every point of each packaged front is the image of a design with every distance variable at 0.5 (g = 0) within 1e-9; the points satisfy sum(f) = (M - 1) / 2 (IDTLZ1) or |1 - f| = 1 (IDTLZ2) within 1e-12.
  • Registry: PlatEMO's dimensions (n = M + 4, M + 9) with 3, 5 and 8 objectives, bounds, encoding, the n_obj override, the warning for a non-standard k, the error for n_var < n_obj, the resolved fronts (3D by default, 2D, none for 4), and an NSGA-II smoke run on both.

tests/reference/test_reference_fronts.py: the four files lie on their analytic fronts within 1e-9 and have at least 1000 points, the mutual non-dominance test covers them, and the resolver test checks their names. The generic registry tests (test_problem_registry.py: default dimensions, the optimize contract) cover the two new keys.

Compatibility

Additive: two new registered problems and four packaged files. No existing problem, front, default or public export changes.

Validation

Run from the worktree with PYTHONPATH set to its src (vamos.__file__ checked):

  • python -m pytest -q -n 4 tests/foundation/test_idtlz_problems.py tests/reference/test_reference_fronts.py tests/foundation/test_problem_registry.py tests/foundation/test_problem_zoo.py tests/foundation/test_problem_evaluation_edge_cases.py tests/engine/test_algorithm_problem_matrix.py tests/foundation/test_problem_reference_values.py tests/ux/test_studio_catalog.py: 2016 passed, 18 skipped (existing skips).
  • ruff check and ruff format --check on the changed Python: clean. python -m compileall: clean. git diff --check: clean.
  • python tools/typecheck.py --scope strict and --scope full: 0 diagnostics; ratchet new=0, increased=0.
  • python tools/check_agent_docs.py: passed. python tools/check_repository_hygiene.py with the new files staged: passed.
  • python -m pytest -m "not slow" -n 4 -p no:cacheprovider -rfE: 10883 passed, 253 skipped, 9 xfailed, 1 failed, tests/experiment/test_cli_ablation.py::test_cli_ablation_runs_and_writes_variants, a known failure under xdist on Windows that also fails on untouched main and passes alone.
  • python tools/health.py: all health checks passed.
  • mkdocs build --strict: built.
  • Generation: python experiments/scripts/generate_reference_fronts.py idtlz1.2d idtlz1.3d idtlz2.2d idtlz2.3d.

Changelog entry

Added (under [Unreleased])

  • Problems: IDTLZ1 and IDTLZ2, the inverted DTLZ1 and DTLZ2 of Jain and Deb (2014) as PlatEMO defines them, registered as idtlz1 and idtlz2 with a configurable number of objectives (n_var = n_obj + 4 and n_obj + 9 by default), with reference fronts for two and three objectives (idtlz1.2d.csv to idtlz2.3d.csv).

Checklist

Architecture health

  • I read the canonical rules in docs/dev/architecture_health.md relevant to this change.
  • I ran the local health command: python tools/health.py.
  • I updated docs/dev/architecture_health.md if I changed guardrails. (No guardrail changed.)

Public API & dependencies

  • Public API changes are intentional; snapshot updated via python tools/update_public_api_snapshot.py if needed. (No public export changes: the problems are registered by name.)
  • No optional/heavy dependencies were added to core [project].dependencies.

Behavior & tests

  • Runtime behavior is unchanged (or explicitly documented as a bugfix). (Two new problems only.)
  • Tests pass locally (fast tier, apart from the known Windows/xdist ablation failure noted above).

Changelog

  • CHANGELOG.md is unchanged, and the changelog entry is under a "Changelog entry" heading in this description.

Reports/retention

  • Raw audit/Goal evidence is external or a CI artifact; only durable conclusions are added to maintained docs.
  • Generated output follows docs/dev/repository_hygiene.md and python tools/check_repository_hygiene.py passes.

🤖 Generated with Claude Code

… them, with reference fronts

The inverted DTLZ1 and DTLZ2 (Jain and Deb, IEEE TEVC 18(4), 2014) are
the usual test of how decomposition-based algorithms cope with an
inverted front; VAMOS had neither (change request P3). They follow
PlatEMO's IDTLZ1.m and IDTLZ2.m, registered as idtlz1 and idtlz2 in the
DTLZ family, with n_obj configurable as for DTLZ:

- IDTLZ1: f = (1 + g) / 2 - 0.5 (1 + g) s, with DTLZ1's g over the last
  k = n - M + 1 variables and its position shape s; n_var defaults to
  n_obj + 4, PlatEMO's D.
- IDTLZ2: f = (1 + g) - (1 + g) s, with g = sum (x_i - 0.5)^2 and
  DTLZ2's shape; n_var defaults to n_obj + 9.

The shape is PlatEMO's fliplr(cumprod([1, h])) .* [1, fliplr(t)], one
helper for both. pymoo 0.6.2's InvertedDTLZ1 is the same function; pymoo
has no inverted DTLZ2. The classes are not exported by vamos.problems.

generate_reference_fronts.py writes idtlz1.2d, idtlz1.3d, idtlz2.2d
and idtlz2.3d from PlatEMO's GetOptimum, (1 - w) / 2 and
1 - w / norm(w) for w on the unit simplex: 1000 points evenly spaced
with two objectives (IDTLZ1's is DTLZ1's segment), and the 5050-point
lattice of the DTLZ samples with three.

Tests: the objectives equal PlatEMO's formulas, written out one
objective at a time in the test, at fixed points with 2, 3 and 5
objectives, and (1 + g) / 2 - DTLZ1 and (1 + g) - DTLZ2; every point of
the fronts is reached by a design with g = 0 and lies on its analytic
front; registry dimensions, warnings, resolved fronts and an NSGA-II
smoke run.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@chatgpt-codex-connector

chatgpt-codex-connector Bot commented Nov 19, 2026 •

Copy link
Copy Markdown
Codex Review Summary: ✅ Completed

This comment shows the latest Codex review activity on this pull request.

Review Status Commit Review trigger
📝 Code Review ✅ Completed 2026-11-19T00:13:39.316423Z 39d048e PR opened
ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

  • Open a pull request for review
  • Mark a draft as ready
  • Comment "@codex review" or "@codex security review".

Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings.

@github-actions

github-actions Bot commented Nov 19, 2026 •

Copy link
Copy Markdown
Contributor

VAMOS documentation preview

https://pr-366-vamos-docs-preview.vamos-optimization.workers.dev

Built from b15a00c. Canonical metadata points to the production .org domain; the preview itself remains isolated.

@NicolasRodriguezUribe
NicolasRodriguezUribe added this pull request to the merge queue Nov 19, 2026
@github-merge-queue
github-merge-queue Bot removed this pull request from the merge queue due to a conflict with the base branch Nov 19, 2026
Keeps both sides of #362 (DTLZ fronts with 4, 6 and 8 objectives) and of
this branch (IDTLZ1 and IDTLZ2): the DTLZ rows of the problem table with
their new fronts and the IDTLZ rows, both provenance rows, both generator
functions and entries, and both sets of front residual checks.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@NicolasRodriguezUribe
NicolasRodriguezUribe added this pull request to the merge queue Nov 19, 2026
Merged via the queue into main with commit 2f3fdbb Nov 19, 2026
15 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant