v0.16.0

PortfolioOptimisers v0.16.0

Diff since v0.15.0

Breaking changes

Old	New	Scope
ReturnRiskMeasure	ExpectedReturn	Type rename
ReturnRiskRatioRiskMeasure	ExpectedReturnRiskRatio	Type rename
Field clr	cle	HierarchicalOptimiser
Field pl (estimators)	ple	All JuMP-based optimisers
Field pl (results)	plr	All JuMP-based optimiser results
Field ple in FactorRiskContribution	frc_ple	FactorRiskContribution
Fields lcs, ct, gcard, sgcard (estimators)	lcse, cte, gcarde, sgcarde	JuMP optimisers
Fields lcs, ct, gcard, sgcard (results)	lcsr, ctr, gcardr, sgcardr	JuMP results

Bug fixes

• UncertaintySetVariance.factory nil-dereference fix when pr === nothing

New features

• Cross-validation framework: KFold, Combinatorial (CPCV), WalkForward (Index and date based variants), MultipleRandomised
• Cross-validation infrastructure: cross_val_predict, predict, fit_predict, OptimisationCrossValidation, NearestQuantilePrediction, PredictionResult, MultiPeriodPredictionResult, PopulationPredictionResult, PredictionReturnsResult
• needs_previous_weights / factory(opt, w) pattern propagated to all optimisation estimators to support weight continuity across CV folds
• is_time_dependent / update_time_dependent_estimator hooks for time-varying constraints
• Index-window (VecInt) parameter on SimpleExpectedReturns and GeneralCovariance for sub-period estimation
• returns_result_view on ReturnsResult for time-windowed slices

Tests

• Tests for new features.

Docs

• New example (examples/10_CrossValidation.jl) and API docs

Merged pull requests:

• Add cross validation and cross validation accessories, and bug fixes. (#69) (@dcelisgarza)
• Improve factory functions fix documentation issues. (#73) (@dcelisgarza)
• Disambiguate between parameters which cannot be result types when used in cross validation and as external estimators in nested clusters and stacking optimisers, and other parameters. (#74) (@dcelisgarza)
• Summarise PR #69 additions and breaking changes (#75) (@Copilot)

Closed issues:

• New feature: Cross Validation. (#38)
• New feature: Non-hierarchical clustering. (#39)
• Cross validation: Walk forward. (#43)
• Cross validation: Multiple randomised. (#44)
• Cross validation: Combinatorial. (#45)

v0.15.0

PortfolioOptimisers v0.15.0

Diff since v0.14.2

Breaking changes

• NOCTracking -> L1Tracking.

• SOCTracking -> L2Tracking.

• SquaredSOCTracking -> SquaredL2Tracking.

New features

• Lp via LpRegularisation and LInf regularisation which respectively use the p- and inf-norm to penalise the objective function.

• Lp and LInf tracking algorithms via LpTracking and LInfTracking which use the p- and inf-norm to track the performance or risk of a portfolio.

Tests

• Added tests for the new features.

Docs

• New docs landing page with new logo and taglines.

• Regularisation example.

• Reduced redundancy and more standardisation of docs.

Maintenance

• Remove double qualification of StatsBase.

• Remove dead code.

• Add newer versions to project compats.

Merged pull requests:

• New logo, code and docs cleanup, new tracking and regularisation features. (#66) (@dcelisgarza)

v0.14.2

PortfolioOptimisers v0.14.2

Diff since v0.14.1

Bug fixes

• Fixed a critical bug where HighOrderPrior was computing coskewness and cokurtosis from the adjusted returns matrix (pr.X) instead of the original returns matrix (X), causing incorrect higher-order moment calculations

Docs

• Added detailed interface documentation with examples for AbstractCovarianceEstimator, AbstractVarianceEstimator, AbstractMatrixProcessingEstimator, AbstractMatrixProcessingAlgorithm, AbstractPosdefEstimator, AbstractDetoneEstimator, and AbstractDenoiseEstimator

• Added two comprehensive examples "Risk Factor Optimisation" and "Improving Moment Estimation"

• Improved front facing and docs readmes.

Maintenance

• Refactored matrix processing, denoise, detone, and posdef functions to return the modified matrix instead of nothing, improving consistency and usability

• Replaced sort! with sort in OWA risk measures to avoid mutating input vectors

• Made the argument checks for linear constraint generation and Black Litterman view generation internally consistent.

Merged pull requests:

• Fix bug in high order prior. Add more docs. Arrive at the standard for interfaces docs. (#65) (@dcelisgarza)

v0.14.1

PortfolioOptimisers v0.14.1

Diff since v0.14.0

Bug fixes

• Fix subtle bug where fees were not being applied when predicting the outer returns of Stacking and NestedClusters optimisations

Docs

• Fix title cases and remove Documenter.jl link disambiguations from repo's readme

Maintenance

• Throw clearer IsNothingError using argcheck rather than letting the program throw an obscure MethodError

Merged pull requests:

• Prepare for cross-validation, fix bug of fees not being used whe computing the outer returns in stacking and nco optimisations. (#64) (@dcelisgarza)

v0.14.0

PortfolioOptimisers v0.14.0

Diff since v0.13.0

Breaking changes

• detone -> dt, denoise -> dn

• plg -> pl

• cent -> ct

• ne -> nt

• nt -> pl

• pe -> pl

• JuMPOptimiser pe -> pr

• pe -> pr for all optimisers

• pe -> pr for all optimisers

• cle -> clr for all optimisers

• BuyInThreshold -> Threshold

• BuyInThresholdEstimator -> ThresholdEstimator

• te -> tr

• EQ -> ==

• LEQ -> <=

• GEQ -> >=

• BrownianDistanceVariance alg -> alg1, algc -> alg2

• NoOptimisationRiskMeasure -> NonOptimisationRiskMeasure

• SquareRelativeErrorWeightFinaliser -> SquaredRelativeErrorWeightFinaliser

• SquareAbsoluteErrorWeightFinaliser -> SquaredAbsoluteErrorWeightFinaliser

• RatioRiskMeasure -> ReturnRiskRatioRiskMeasure

• GreedyAllocationOptimisation -> GreedyAllocationResult

• DiscreteAllocationOptimisation -> DiscreteAllocationResult

Tests

• Increase test coverage

Docs

• Update readme

• Update docs

• Remove empty docstrings, it's better to use the auto generated struct info

Maintenance

• Configure Copilot instructions following GitHub best practices (#61)

• Simplify factory functions

• Dictionary of standard terms

• Make optimisation estimator structs' orders consistent

• Make API consistent for a lot of functions

Merged pull requests:

• Configure Copilot instructions following GitHub best practices (#61) (@Copilot)
• Stabilising the api. (#63) (@dcelisgarza)

Closed issues:

• ✨ Set up Copilot instructions (#60)

v0.13.0

PortfolioOptimisers v0.13.0

Diff since v0.12.1

Breaking Changes

Parameter Renames (consistency and brevity)

• sce → sca, scei → scai, sceo → scao (scalariser variants)
• executor → ex
• target → tgt
• threshold → t

Type Renames

• ClusteringEstimator → ClustersEstimator
• EllipseUncertaintySetAlgorithm → EllipsoidalUncertaintySetAlgorithm
• UniformlyDistributedBounds → UniformValues and simplified type hierarchy
• KellyReturn → LogarithmicReturn
• Optimization result types made consistent

New features

• Cholesky caching: Added chol field to structs benefiting from pre-computed decompositions
• High order factor prior: Support for additional factor priors with residual variance validation
• Missing factory/view functions: Added previously absent constructor and accessor functions
• Numerical stability: Improved HERC log-sum-exp scalariser stability
• K-means clustering: Other than hierarchical optimisations, other features only need knowledge of cluster assignment rather than a full phylogeny, this implements k-means clustering for such cases

Maintenance

• Main module restructured from bulk using to explicit imports
• Increased test coverage
• Improve type stability of some view functions
• Keep improving documentation

Merged pull requests:

• Dev (#54) (@dcelisgarza)
• Breaking changes, high order factor prior, add chol field to structs that can benefit from it, add missing factory and view functions. (#55) (@dcelisgarza)

v0.12.1

PortfolioOptimisers v0.12.1

Diff since v0.11.1

Bug fixes

• Correctness fix to drawdowns auxiliary function used for plotting.

New features

• Add Distributionally Robust Conditional Drawdown at Risk.

• Add Power Norm X at Risk measures.

• Enable Drawdown at Risk for non-clustering optimisation.

Docs

• Docs up to Fees.

Maintenance

• Remove side effects from Y X at Risk risk measure families.

Merged pull requests:

• New covariance estimators, new matrix processing estimators. (#49) (@dcelisgarza)
• Improve mathematical descriptions in docs. Add power norm x at risk measure, add scenario weights to drawdown risk measures. Add distributionally robust drawdown at risk measure. (#53) (@dcelisgarza)

Closed issues:

• JuliaRegistrator (#50)

v0.11.1

PortfolioOptimisers v0.11.1

Diff since v0.11.0

Bug fixes

• Fix bug in prices_to_returns when using a benchmark.

Docs

• Docs up to 03_Preprocessing.

Maintenance

• Update pre-commit.

Merged pull requests:

• Fix API bug in prices_to_returns when using a benchmark. Add more docs. (#48) (@dcelisgarza)

v0.11.0

PortfolioOptimisers v0.11.0

Diff since v0.10.0

Breaking

• Enable the use of cross validation.

• Add default fill values to various estimators.

• Add custom mu override to ArithmeticReturn.

• Update various function APIs.

• Add key to buy in threshold and linear constraint estimators which enable a new class of set cardinality constraints.

New features.

• Add benchmark returns to prices_to_returns.

• Add squared soc tracking.

• Add quad tracking to squared and quad risk measure constants.

Bug fixes

• Fix various type signature bugs.

• Fix Adjust_Risk_Contributions.jl constants.

• Fix API bugs in set group cardinality constraints.

• Enable set group mip constraints.

Maintenance

• Start sprucing up asserts.

• Start adding struct tests.

• Standardise and use type aliases

• Simplify assertions.

• Fix various type instabilities.

• Add missing functions for Nothing.

• Enable more Julia magic by broadening type bounds.

• Add missing unions and functions that should be able to operate on them.

• Don't remove rf*k from ret expression, do it at the level of the objective expression, this is more correct.

• Fix typos.

• Remove unnecessary functions.

• Add missing factory functions, eliminate ambiguities.

• Split format and link checking into separate workflows.

• Refactor optimise function and iv` assignment.

• Add `_optimise function to the base optimisation file.

• Add set group and set cardinality tests.

• Specify read permission to format check.

Documentation

• Update docs and arrive at what they should look like, which is 01_Base.jl and Base.md.

• Update npm.

• Improve doctest setup and filters.

• Programmatically include all symbols for documentation purposes.

• Test docs completeness.

Merged pull requests:

• Type aliases, benchmark returns for prices_to_returns, fix set MIP constraint bugs. (#47) (@dcelisgarza)

v0.10.0

PortfolioOptimisers v0.10.0

Diff since v0.9.1

Breaking changes

• Change ltd covariance name to its full description, increase test coverage.

Bug fixes

• Fix various typing bugs.

• Fix bug where provided weights are not passed to argument in entropy pooling.

• Fix group_to_val! to enable its use in nested clustered optimisations.

• Add missing factory functions for dimensionality reduction regression.

• Fix when using weights in dimensionality reduction regression.

Documentation

• Fix contents in documentation.

Maintenance

• Remove dead code.

• Make some functions private.

• Add missing convenience function, increase coverage.

• Optimise code.

• Replace throws with argcheck.

• Make all argument checks more explicit.

Merged pull requests:

• Incremental improvements. (#34) (@dcelisgarza)