Skip to content

Convert all algorithm documentation to the new format. - #690

Merged
janosg merged 13 commits into
mainfrom
standardize-algo-docs
Nov 15, 2026
Merged

janosg merged 13 commits into
mainfrom
standardize-algo-docs

Conversation

@janosg

@janosg janosg commented Jul 4, 2026 •

Copy link
Copy Markdown
Member

Previously, algorithm documentation was in algorithms.md; now it's in the docstrings of the algorithm classes. The new format is described here. While converting, we found and fixed the issues listed below. Every docstring was checked against the optimagic wrapper code and the upstream sources: scipy 1.17.1, NLopt 2.10.1, pagmo2, Ipopt 3.14 / cyipopt 1.6.1, DFO-LS 1.6.2, Py-BOBYQA 1.5.0, fides 0.7.4 and tranquilo 0.1.1. Where the wrapper behaves unexpectedly, the docstrings describe what it actually does. The underlying code bugs are fixed in follow-up PRs or listed in #583. Apart from corrected error and warning messages, this PR changes no code.

0. Changes to algorithms.md and the how-to guide

  • Option docs did not render at all. Since the switch to pydantic (Use pydantic for algo option validation (instead of our own validation logict). #696), mark.minimizer re-creates every algorithm class as a pydantic dataclass that inherits from the original class. Autodoc therefore treated all documented fields as inherited and hid them. Every autoclass now uses :inherited-members: Algorithm, object, as the gfo and pyswarms entries already did. The how-to guide explains why.
  • Copy-pasteable examples (review comment). Every example passes params as a NumPy array, has no ... placeholders and is self-contained. Bounds are added where the algorithm needs them, and least-squares and likelihood examples mark their functions. All 184 snippets were executed. Only nevergrad_bo (needs bayes-optim) and tao_pounders (Linux only) could not be run. This also fixed examples that were broken before:
    • bayes_opt used a nonexistent n_iter option.
    • pyswarms used nonexistent topology_type and k_neighbors options.
    • The iminuit example had a syntax error.
    • The shgo example crashed without bounds.
    • In the nag_pybobyqa string-interface example the option was silently dropped (see code issues).
  • Fixed typos and errors in the intros: the contradictory fides version note ("fides>=0.7.4, make sure you have at least 0.7.1"), "AX-platfofm" and "Toolkit of Advanced optimization".
  • refs.bib:
    • Fixed the garbled author fields of Thomas2000 (Runarsson & Yao) and DaSilva2010/DaSilva2010a.
    • Corrected the title of Dembo1983.
    • Added ChiangZavala2016 for the Ipopt neg_curv_test_* options. They cited an unrelated paper before.

1. Errors in the old documentation (corrected)

scipy

  • Stale constraint-support notes. The old text said "SLSQP's / COBYLA's / trust_constr's general nonlinear constraints are not supported yet by optimagic". All three have supports_nonlinear_constraints=True today. The new docstrings instead explain how optimagic converts equality constraints to pairs of inequality constraints for COBYLA. COBYLA natively only handles inequalities.

  • Copy-paste error in scipy_newton_cg. The warning advised refining "an optimum found with Powell" with another optimizer. It meant an optimum found with newton_cg.

  • Wrong default for scipy_direct. The old text claimed that the volume tolerance eps defaults to 1e-6, "differing from scipy's default 1e-4". The actual optimagic default is CONVERGENCE_FTOL_REL = 2e-9. A smaller eps makes the search more local, not a larger one.

  • Wrong Powell convergence formula. The documented ftol formula was the one from L-BFGS-B. Powell uses 2(f^k - f^{k+1}) <= ftol (|f^k| + |f^{k+1}|) + 1e-20. xtol only controls the precision of the line search; it is not a stopping criterion.

  • Nonexistent options documented. scipy_truncated_newton was documented with stopping.maxiter, but the class only has stopping_maxfun. scipy_ls_lm was documented with tr_solver/tr_solver_options, which only exist for the trf and dogbox least-squares solvers. Both were dropped.

  • Misleading option semantics:

    • COBYLA's stopping_maxiter limits function evaluations.
    • Newton-CG's xtol is an average absolute step criterion.
    • BFGS uses the strong Wolfe curvature condition.
    • basinhopping runs n_local_optimizations local optimizations in total (SciPy niter = n - 1).
    • shgo's minhgrd stops after a single iteration whose growth is at most the given value.
    • brute's finish is called as finish(func, x0, args=...).
    • trust_constr uses the interior point method whenever bounds or nonlinear constraints are present.
  • shgo:

    • SciPy's default local method is SLSQP, not L-BFGS-B.
    • trust-constr also uses the constraints.
    • Custom local functions are not supported.
    • Bounds are required.
    • n_simplex_iterations has no effect because maxev is always passed.
  • Outdated implementation claims. COBYLA has used PRIMA's pure-Python implementation since SciPy 1.16. L-BFGS-B has used a C translation since SciPy 1.15.

  • Several places where optimagic's defaults silently differ from SciPy are now stated explicitly:

    Algorithm Option optimagic SciPy
    Nelder-Mead ftol 1e-8 1e-4
    SLSQP ftol 1e-8 1e-6
    BFGS xrtol 1e-5 0
    trust_constr xtol 1e-5 1e-8
    dual_annealing maxfun 1e6 1e7
    differential_evolution atol 1e-8 0
    TNC xtol 0 sqrt(eps)
    TNC gtol 1e-5 1e-2·sqrt(accuracy)
    COBYLA, trust_constr initial radius 0.1·max(‖x0‖∞, 1) 1
    shgo local method L-BFGS-B SLSQP
    shgo n 128 100
    shgo maxev 1000 None
    brute finish None fmin
  • Typos fixed: finitie_difference_precision and messg_num → mesg_num.

nlopt

  • MMA/CCSAQ constraint support was backwards. The old text said these support "nonlinear equality (but not inequality) constraints". NLopt supports the opposite: nonlinear inequality constraints, not equality constraints. In optimagic, MMA gets equalities via conversion to inequality pairs. CCSAQ currently exposes no nonlinear constraints at all (supports_nonlinear_constraints=False).
  • xtol_rel semantics were outdated. NLopt uses an L1-norm test, sum |Δx| < tol · sum |x|, not a per-parameter test.
    • BOBYQA, COBYLA and NEWUOA apply it to the trust-region radius.
    • DIRECT applies it to the rectangle size relative to the bound width.
    • ESCH ignores all convergence criteria and always uses stopping_maxfun.
  • PRAXIS:
    • The claim that optimagic disables bounds for PRAXIS was wrong. Bounds are forwarded to NLopt, which has enforced them via a variable transformation since NLopt 2.9.1.
    • The outdated "returns infinity" and "significantly reduced speed" remarks were removed.
    • Seeding does work, via nlopt.srand.
  • DIRECT:
    • The old variant list offered a nonexistent "DIRECT_RAND" and omitted "DIRECT_NOSCAL".
    • The rationale for NOSCAL was reversed.
    • random_search=True requires locally_biased=True; other combinations currently raise an error.
    • The start parameters are not used.
  • Wrong algorithm details:
    • The dropdown was titled nlopt_lbfgs, but the algorithm is registered as nlopt_lbfgsb.
    • nlopt_tnewton is the plain LD_TNEWTON, not "preconditioned".
    • COBYLA is the only derivative-free local NLopt optimizer with nonlinear constraints; ISRES is a global one.
    • NEWUOA ignores one-sided bounds; there is now a warning.
  • Added NLopt facts:
    • The CRS population must be at least n+1.
    • TNewton, LBFGSB and VAR have a hidden gradient tolerance of 1e-8.
    • The NLopt author recommends CCSAQ over MMA.
    • Nelder-Mead applies no ftol_rel.
  • "Globally convergent" for MMA/CCSAQ is clarified per NLopt's own note: convergence to a local optimum from any feasible starting point.
  • Removed an unsourced "<10 past updates" claim for LBFGS storage.
  • Typos fixed (Nedler-Mead, alggorithm, faunction, a :cite: role missing its colon).

pygmo

  • Wrong default population size formula. The old text said "twice the number of parameters but at least X". The code (get_population_size) uses 10 * (n_params + 1), clipped below at the algorithm-specific minimum.

  • ftol/xtol descriptions swapped for sade, de1220, cmaes and xnes. This is a known pagmo documentation bug that had been copied into algorithms.md.

  • Wrong option ranges and defaults, checked against the pagmo C++ source:

    • de.weight_coefficient is in [0, 1], not [0, 2].
    • sga.mutation_polynomial_distribution_index is in [1, 100].
    • sga.selection_tournament_size defaults to 2.
    • compass_search.stop_range must be strictly smaller than start_range.
    • Several ranges are open at 0.
  • Wrong behavior claims:

    • Integer codes for variants are no longer accepted, because pydantic enforces the Literal types.
    • seed only seeds the initial population.
    • gaco's stopping_maxfun is pagmo's evalstop, which counts generations without improvement.
    • IHS does not support stochastic problems.
    • The signs of the PSO velocity formula were reversed.
    • mbh's default inner algorithm is compass search with max_fevals=1.
    • The cmaes and compass_search step sizes are relative to the bound width.
    • simulated_annealing and compass_search only use the population to choose a starting point.
  • Undocumented deviations from pygmo defaults are now stated:

    Algorithm Option optimagic pygmo
    bee_colony max_n_trials 1 20
    mbh stopping_max_inner_runs_without_improvement 30 5
    simulated_annealing end temperature 0.01 0.1
    simulated_annealing n_range_adjustments 10 1
    simulated_annealing bin_size 10 20
    xnes initial_search_share 1 0.5
  • A broken .. note: directive is fixed.

  • GWO: fixed typos, and the criticism of the algorithm is attributed to the pagmo developers.

  • The batch_evaluator bullets for gaco and pso_gen were dropped; these classes only have n_cores.

ipopt

  • Contradictory linear_solver default. The old text said the default is "ma27", while its own value list said "mumps (default)". The code default is "mumps"; Ipopt's own documented default is "ma27".

  • Options are always passed explicitly. This turns off Ipopt's behavior for unset options, and the affected docstrings now say so:

    • mu_max_fact has no effect.
    • mu_min is used as given.
    • recalc_y and mu_strategy="adaptive" are not switched on for limited-memory.
    • slack_bound_* does not follow bound_*.
    • expect_infeasible_problem does not tighten required_infeasibility_reduction.
  • Unusable or surprising options now carry warnings:

    • skip_finalize_solution_call returns the start parameters with success=True.
    • mehrotra_algorithm aborts unless four further options are set.
    • hessian_approximation="exact" runs without a valid Hessian, because optimagic passes none.
    • warm_start_same_structure aborts.
  • Other options documented with their actual behavior:

    • Warm starts begin with zero multipliers.
    • warm_start_entire_iterate has no effect.
    • "user-scaling" means no scaling.
    • A negative obj_scaling_factor cannot replace om.maximize.
  • Default mismatches:

    • acceptable_dual_inf_tol: the old text documented Ipopt's default of 1e+10, but the code passes 1e-10 (see code issues). The docstring now documents the actual behavior, including that acceptable termination effectively never triggers.
    • resto_failure_feasibility_threshold: the old text said 0; the wrapper maps None to 100 * convergence_ftol_rel.
  • Wrong references.

    • alpha_min_frac is Eqn. (23), not (20).
    • The neg_curv_test_* options cited the wrong paper.
    • A nonexistent "kkt-error-filter" option was referenced.
    • The square-problem auto-enable of expect_infeasible_problem is commented out in Ipopt 3.14.
  • Mangled bullets fixed. theta_max_fact had been fused into the watchdog_trial_iter_max bullet, and recalc_y_feas_tol was mislabeled.

  • Value typos:

    • "10-05" → 1e-5 (filter_margin_fact)
    • "10-06" → 1e-6 (sigma_min)
    • "bound_mult" → "bound-mult"
    • adaptive_mu_kkterror_red_iters is an int, not a float
  • Deviations from Ipopt defaults are now stated:

    Option optimagic Ipopt
    tol 2e-9 1e-8
    max_iter 1,000,000 3000
    hessian_approximation "limited-memory" "exact"
    option_file_name "" "ipopt.opt"

    The class docstring names the renamed options and lists all options that are not exposed.

NAG (DFO-LS, Py-BOBYQA)

  • seek_global_optimum restriction was wrong. The old text said: "Only applies for noisy criterion functions." Py-BOBYQA's documentation demonstrates it on noise-free functions. The actual requirement is that lower and upper bounds are both specified (individual entries may be infinite).
  • Fast-start default condition was inverted. DFO-LS uses "jacobian" when the number of residuals is at least the number of parameters. In optimagic, "auto" currently always behaves like "jacobian" (see code issues); this is now documented on the option.
  • Wrong defaults and semantics:
    • scale_of_trustregion_step_perturbation is always 1.
    • The fast-start shrinking_factor_not_successful uses DFO-LS's default, not the user's value.
    • The advice to set npt to len(x) was dropped, because DFO-LS requires npt ≥ len(x)+1.
    • floor_of_jacobian_singular_values has no effect.
    • The Py-BOBYQA use_resets default also depends on seek_global_optimum.
    • The n_extra_* reset options and the auto-detect criteria are corrected.
    • Py-BOBYQA's slow-progress test is absolute, not relative.
    • DFO-LS also stops at f ≤ 1e-12 (model.abs_tol).
  • Wrong option value in prose. "trustregion_step" should be "trustregion", the value the code accepts.
  • "root contributions" is replaced by "residuals".
  • noise_n_evals_per_point is documented with its actual positional calling convention, plus a warning that it is currently unusable.
  • Module-level constant docstrings that duplicated and contradicted the class docs were removed.
  • Stale :ref: pointers were removed.

fides

  • Wrong documented defaults.
    • trustregion.subspace_dimension was documented as defaulting to "2D". The optimagic default is "full"; the fides package itself defaults to "2D".
    • optimagic defaults stepback_strategy to "truncate", while fides defaults to "reflect".
    • The fatol, frtol, gatol, grtol and maxiter defaults also differ from fides; this is now stated.
  • Wrong trust-region parameter descriptions.
    • mu is not an acceptance threshold. fides accepts every step with a positive ratio, and mu only triggers a radius decrease.
    • mu, eta, γ1 and γ2 are fides' option names, not symbols from Nocedal & Wright's Algorithm 4.1. Only their default values come from there.
  • Nonexistent options dropped. trustregion.refine_stepback and trustregion.scaled_gradient_as_possible_stepback were documented but are not fields of the class.
  • The convergence criteria formulas were verified against fides 0.7.4 (Optimizer.check_convergence).
  • The claim that fides solves the subproblem "exactly" now only applies to subspace_dimension="full".
  • The install hint is quoted, so the shell does not treat > as a redirect.

TAO pounders

  • Outdated link. The old text linked github.com/erdc/petsc4py, an unofficial and outdated mirror.
  • Wrong disabling semantics. Zero (or another falsy value) is passed to TAO as -1, which restores TAO's default (1e-8 for gatol and grtol). Only gttol ends up disabled.
  • Inactive gtol criteria. optimagic always installs a custom maxiter convergence test, which replaces TAO's default test. All convergence_gtol_* criteria therefore currently have no effect, and reaching stopping_maxiter is reported as success. This is now documented.
  • LaTeX typos (missing backslash on \epsilon, X0 → X_0).

Own optimizers (bhhh, neldermead_parallel, pounders)

  • bhhh:
    • The outdated interface description (fun returning a dict with a "contributions" entry) is replaced by the current @om.mark.likelihood interface.
    • New warning: the implementation steps along +H⁻¹g, so with derivatives consistent with fun it moves uphill (see code issues).
  • neldermead_parallel:
    • The class docstring was a broken numpydoc stub that documented nonexistent criterion/x parameters.
    • adaptive adapts to the problem dimension (Gao–Han).
    • The n_cores cap only applies to parallel reflections.
    • xtol uses the infinity norm.
  • pounders:
    • The gtol_rel default was written as "1-8".
    • trustregion_expansion_factor_successful was described as a "Shrinking factor".
    • The radius-update conditions are corrected: expansion requires rho ≥ threshold and a long step, and accepted steps can shrink the radius.
    • "steihaug-toint" → "steihaug_toint".
    • The options are documented under their current names.
    • The nonexistent batch_evaluator option was dropped.
    • The gtol docs are consistent with Fix gradient norm scaling in pounders convergence checks #708.

Other documentation fixes

  • nevergrad_wizard / nevergrad_portfolio dropdowns were broken. They referenced classes that no longer exist (NevergradWizard, Wizard, NevergradPortfolio, Portfolio). The dropdowns are renamed to nevergrad_ngopt / nevergrad_meta, and the examples are rewritten.
  • how_to_add_optimizers.ipynb crashed the docs build. Its @mark.minimizer call was missing the now-required needs_bounds and supports_infinite_bounds arguments. Also fixed a dead cross-reference and typos.
  • tranquilo was not documented anywhere. It is now documented from scratch, based on the tranquilo paper (Gabler, Gsell, Mensinger, Petrosyan 2024, CRC TR 224 DP 522) and verified against the tranquilo 0.1.1 source. The docstrings state where 0.1.1 behaves unexpectedly:
    • Candidates are accepted on any improvement.
    • The parallel line search only runs with acceptance_decider="classic_line_search".
    • Noisy or parallel optimization of scalar functions is flagged as experimental by tranquilo.
    • model_type is fixed per version.
    • stopping_maxtime and convergence_min_trust_region_radius have no effect.
    • seed=None raises an error.
    • batch_evaluator is not forwarded.

🤖 Generated with Claude Code

@review-notebook-app

Copy link
Copy Markdown

Check out this pull request on  ReviewNB

See visual diffs & provide feedback on Jupyter Notebooks.


Powered by ReviewNB

@codecov

codecov Bot commented Jul 4, 2026 •

Copy link
Copy Markdown

Codecov Report

❌ Patch coverage is 99.72603% with 2 lines in your changes missing coverage. Please review.

Files with missing lines Patch % Lines
src/optimagic/optimizers/nag_optimizers.py 98.18% 1 Missing ⚠️
src/optimagic/optimizers/pygmo_optimizers.py 99.32% 1 Missing ⚠️
Files with missing lines Coverage Δ
src/optimagic/optimizers/bhhh.py 96.36% <100.00%> (+0.20%) ⬆️
src/optimagic/optimizers/fides.py 96.77% <100.00%> (+0.72%) ⬆️
src/optimagic/optimizers/ipopt.py 98.48% <100.00%> (+1.12%) ⬆️
src/optimagic/optimizers/neldermead.py 92.74% <100.00%> (+0.50%) ⬆️
src/optimagic/optimizers/nlopt_optimizers.py 97.57% <100.00%> (+0.77%) ⬆️
src/optimagic/optimizers/pounders.py 93.37% <100.00%> (+0.90%) ⬆️
src/optimagic/optimizers/scipy_optimizers.py 97.75% <100.00%> (+0.78%) ⬆️
src/optimagic/optimizers/tao_optimizers.py 76.52% <100.00%> (+1.29%) ⬆️
src/optimagic/optimizers/tranquilo.py 99.51% <100.00%> (+0.30%) ⬆️
src/optimagic/optimizers/nag_optimizers.py 94.26% <98.18%> (+1.02%) ⬆️
... and 1 more

... and 2 files with indirect coverage changes

🚀 New features to boost your workflow:
  • ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

janosg and others added 2 commits July 5, 2026 17:13
… annotations.

With `from __future__ import annotations`, dataclass field types are
annotation strings rather than type objects, so the TYPE_CONVERTERS lookup
in Algorithm.__post_init__ never matched and option coercion/validation was
silently skipped for all optimizer modules using the import.

Adds a string-keyed TYPE_CONVERTERS_BY_NAME lookup so coercion works
identically with and without the import, plus regression tests: a dummy
algorithm defined under postponed annotations mirroring the existing type
conversion tests, and a registry-wide test that passes floats for all
int-typed options of every algorithm.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
@janosg
janosg force-pushed the standardize-algo-docs branch from 51d4bec to cde6d90 Compare July 5, 2026 15:15
@janosg
janosg changed the base branch from main to fix-algo-option-type-conversion July 5, 2026 15:15
@janosg
janosg changed the base branch from fix-algo-option-type-conversion to main August 3, 2026 08:25
janosg and others added 2 commits August 3, 2026 10:25
Brings in #696, which replaces the hand-written option converter table with
pydantic validation. PR #691, whose fix for the same problem was the first
commit on this branch, was closed in favor of #696, so its changes are dropped
here rather than merged:

- src/optimagic/type_conversion.py: deleted (superseded by validated_dataclass
  in optimagic/typing.py), including the TYPE_CONVERTERS_BY_NAME lookup #691
  added.
- Algorithm.__post_init__: removed, main validates via mark.minimizer instead.
- tests/.../test_algorithm_future_annotations.py: deleted. Its coverage of the
  postponed-annotations path is provided by main's
  test_algo_option_validation_stringified.py, and its registry-wide int
  coercion test by test_all_algorithms_can_be_instantiated_with_defaults.

Docs conflicts resolved by keeping this branch's docstrings on top of main's
widened annotations:

- Fides.hessian_update_strategy is now HessianUpdateStrategy, not a bare
  Literal, so its docstring no longer claims that HessianApproximation
  instances are unavailable; it explains that Broyden must be passed as an
  instance because phi has no default.
- PygmoMbh.inner_algorithm uses the PygmoAlgorithm alias.

Also documents the new constraint that field annotations must be resolvable at
runtime in how_to_document_optimizers.md, since pydantic now resolves them and
a TYPE_CHECKING-only name in a field type raises on instantiation.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Comment thread CONSTRAINTS_REFACTOR_REVIEW.md Resolved
Comment thread CLEANUP_NOTES.MD Resolved
Comment thread REPORT.md Resolved
Comment thread docs/source/algorithms.md

Copy link
Copy Markdown
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

For the algorithm examples, the value passed to params parameter is a list, and the lambda function passed to fun does a matrix multiplication. If someone tries to directly copy and paste the example, it fails. How about changing the lambda function to lambda x: sum(v**2 for v in x) or the list into a NumPy array, and also removing the ellipsis from the algorithm options?


For example:

import optimagic as om

om.minimize(
    fun=lambda x: sum(v**2 for v in x),
    params=[1.0, 2.0, 3.0],
    algorithm=om.algos.scipy_neldermead(stopping_maxiter=1_000),
)

Or

import numpy as np
import optimagic as om

om.minimize(
    fun=lambda x: x @ x,
    params=np.array([1.0, 2.0, 3.0]),
    algorithm=om.algos.scipy_neldermead(stopping_maxiter=1_000),
)

janosg and others added 9 commits September 29, 2026 12:13
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Removes CLEANUP_NOTES.MD, CONSTRAINTS_REFACTOR_REVIEW.md and REPORT.md, which were
committed by accident, and fixes "libray" and "simple have a look" in
how_to_add_optimizers.ipynb.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Every example in algorithms.md now passes params as a NumPy array, drops the `...`
placeholders and is self-contained, so each code block runs as is. Bounds are added
for all algorithms that need them, and least-squares and likelihood examples mark
their functions accordingly. This also fixes examples that used options that don't
exist (bayes_opt `n_iter`, pyswarms `topology_type` and `k_neighbors`), a syntax
error in the iminuit example, a shgo example that crashed without bounds, and a
nag_pybobyqa string-interface example whose option was silently dropped.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Since mark.minimizer re-creates algorithm classes as pydantic dataclasses that
inherit from the original class, autodoc treated all documented fields as inherited
and did not show them. Every autoclass directive in algorithms.md now uses
`:inherited-members: Algorithm, object`, as the gfo and pyswarms entries already did,
and the how-to guide explains why. The guide's quote of the ScipyLBFGSB docstring is
synced with the corrected docstring.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…sources.

Every class and field docstring was checked against the optimagic wrapper code
and the upstream sources (scipy 1.17.1, NLopt 2.10.1, pagmo2, Ipopt 3.14 /
cyipopt 1.6.1, DFO-LS 1.6.2, Py-BOBYQA 1.5.0, fides 0.7.4, tranquilo 0.1.1).
Wrong formulas, defaults, option semantics and references are corrected,
undocumented deviations from upstream defaults are stated, and where the wrapper
behaves unexpectedly the docstrings describe the actual behavior. The stale
module-level constant docstrings in nag_optimizers.py, which contradicted the
class docs, are removed. Typos in a few error and warning messages are fixed; no
logic changes. refs.bib gets corrected author fields and titles and a new entry
for the Ipopt negative curvature test.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
# Conflicts:
#	docs/source/how_to/how_to_add_optimizers.ipynb
# Conflicts:
#	docs/source/algorithms.md
#	src/optimagic/optimizers/nlopt_optimizers.py
@janosg
janosg merged commit c8177de into main Nov 15, 2026
22 checks passed
@janosg
janosg deleted the standardize-algo-docs branch November 15, 2026 13:04
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.

3 participants