400 refactor network class to interface by apphp · Pull Request #401 · RubixML/ML

apphp · 2026-03-10T23:41:11Z

No description provided.

* Initial commit * Better testing * Improve the docs * Rename benchmark * Explicitly import max() function

* Add Wrapper interface for models wrappers * Add WrapperAware trait * Fix PhpDoc * Revert "Add WrapperAware trait" This reverts commit 241abc4. * Rename Wrapper interface to EstimatorWrapper * PHP CS fix

* add Swoole backend * phpstan: ignore swoole * feat: swoole process scheduler * fix(swoole): redo tasks when hash collision happens * chore(swoole): make sure coroutines are at the root of the scheduler * chore(swoole): set affinity / bind worker to a specific CPU core * chore(swoole): use igbinary if available * fix: remove comment * fix(swoole): worker cpu affinity * fix(swoole): cpu num * feat: scheduler improvements * style * chore(swoole): remove unnecessary atomics * chore(swoole): php backwards compatibility * fix: phpstan, socket message size * fix: uncomment test * style: composer fix

* Initial commit * Allow deltas in units tests

* add Swoole backend * phpstan: ignore swoole * feat: swoole process scheduler * fix(swoole): redo tasks when hash collision happens * chore(swoole): make sure coroutines are at the root of the scheduler * chore(swoole): set affinity / bind worker to a specific CPU core * chore(swoole): use igbinary if available * fix: remove comment * fix(swoole): worker cpu affinity * fix(swoole): cpu num * feat: scheduler improvements * style * chore(swoole): remove unnecessary atomics * chore(swoole): php backwards compatibility * fix: phpstan, socket message size * fix: uncomment test * style: composer fix * docs: Swoole backend

…ubixML#327)

* chore: add NumPower to workflow pipeline. * chore: install openbla. * chore: install openblas from apt repo.

* Add is_finite checks to skip NAN values * Tests and fix issue with samples that are all non-finite

…o in Swish

…lasses

…atures, num_samples]

…tropy loss function

392 convert layers to numpower

…-to-NumPower 394 convert misc nn classes to num power

Copilot

Pull request overview

This PR upgrades the codebase to a PHP 8.4+/PHPUnit 12 toolchain while refactoring several ML/neural-net components (including network/optimizer/initializer abstractions), updating serialization/backends, and modernizing type declarations across the project.

Changes:

Modernize typing/API surfaces (strict_types, typed properties/returns, named args) and update PHPUnit tests to attributes.
Add/extend neural-net abstractions and implementations (optimizers, initializers, activation + cost functions, snapshot/shape assertions).
Introduce Swoole backend support and change extractor exporting behavior to append-by-default with an overwrite flag.

Reviewed changes

Copilot reviewed 269 out of 584 changed files in this pull request and generated 22 comments.

Show a summary per file

File	Description
tests/Backends/Tasks/TrainLearnerTest.php	Modernize test (attributes, strict types, named args).
tests/Backends/Tasks/TrainAndValidateTest.php	Modernize test (attributes, strict types, named args).
tests/Backends/Tasks/ProbaTest.php	Modernize test (attributes, strict types, named args).
tests/Backends/Tasks/PredictTest.php	Modernize test (attributes, strict types, named args).
tests/Backends/SwooleTest.php	Add Swoole backend test coverage (extension-gated).
tests/Backends/SerialTest.php	Modernize backend tests and task construction.
tests/Backends/AmpTest.php	Modernize backend tests and task construction.
src/functions.php	Tighten iterator_first typing/behavior (now throws on empty).
src/constants.php	Bump library VERSION to major-only string.
src/Transformers/TfIdfTransformer.php	Rename TF-IDF dampening flag to sublinear.
src/Transformers/MinMaxNormalizer.php	Ignore non-finite values in fit/transform/reverseTransform.
src/Transformers/MaxAbsoluteScaler.php	Ignore non-finite values when updating max-abs.
src/Transformers/LambdaFunction.php	Add mixed typing for context parameter.
src/Transformers/IntervalDiscretizer.php	Encode bin ordinals as base-26 strings.
src/Transformers/ImageRotator.php	Minor array formatting / trailing comma fix.
src/Transformers/BooleanConverter.php	Tighten typing and enforce same-type true/false replacements.
src/Traits/AssertsShapes.php	Add shared NDArray shape assertion helper.
src/Strategies/Constant.php	Add union typing and return type.
src/Specifications/SwooleExtensionIsLoaded.php	Add specification to validate swoole extension availability.
src/Specifications/SpecificationChain.php	Remove a stale/incorrect `@throws` annotation.
src/Serializers/RBX.php	Adjust RBX compatibility checks to support format versions range.
src/Regressors/SVR.php	Use null-coalescing assignment and add predictSample return type.
src/Regressors/Ridge.php	Refactor Ridge to NDArray-based math for training/prediction.
src/Regressors/RegressionTree.php	Add predictSample return type.
src/Regressors/RadiusNeighborsRegressor.php	Add predictSample return type.
src/Regressors/KNNRegressor.php	Add predictSample return type.
src/Regressors/KDNeighborsRegressor.php	Add predictSample return type.
src/Regressors/ExtraTreeRegressor.php	Add predictSample return type.
src/Regressors/Adaline.php	Refactor network type to Network interface.
src/Pipeline.php	Add __call() return type mixed.
src/PersistentModel.php	Use null-coalescing assignment and add __call() return type mixed.
src/NeuralNet/Snapshots/Snapshot.php	Add snapshot/restore of parametric layer parameters.
src/NeuralNet/Parameter.php	Add __clone() return type void.
src/NeuralNet/Optimizers/Stochastic/Stochastic.php	Add SGD optimizer implementation under new namespace.
src/NeuralNet/Optimizers/Cyclical.php	Fix typo in exception message.
src/NeuralNet/Optimizers/Base/Optimizer.php	Add Optimizer interface contract.
src/NeuralNet/Optimizers/Base/Adaptive.php	Add Adaptive optimizer contract.
src/NeuralNet/Networks/Base/Contracts/Network.php	Add base network contract for layers traversal.
src/NeuralNet/Network.php	Update attribution in Network interface.
src/NeuralNet/Layers/Binary.php	Add explanatory comment for BCE+sigmoid gradient shortcut.
src/NeuralNet/Layers/Base/Contracts/Parametric.php	Add Parametric layer contract for parameters/restore.
src/NeuralNet/Layers/Base/Contracts/Output.php	Add Output layer contract for backprop endpoint.
src/NeuralNet/Layers/Base/Contracts/Layer.php	Add Layer contract using NDArray forward/infer.
src/NeuralNet/Layers/Base/Contracts/Input.php	Add Input marker interface.
src/NeuralNet/Layers/Base/Contracts/Hidden.php	Add Hidden layer backprop contract.
src/NeuralNet/Initializers/Xavier/XavierUniform.php	Add Xavier uniform initializer.
src/NeuralNet/Initializers/Xavier/XavierNormal.php	Add Xavier normal initializer.
src/NeuralNet/Initializers/Uniform/Uniform.php	Add Uniform initializer and beta validation.
src/NeuralNet/Initializers/Uniform/Exceptions/InvalidBetaException.php	Add custom exception for invalid beta.
src/NeuralNet/Initializers/Normal/TruncatedNormal.php	Add truncated normal initializer with validation.
src/NeuralNet/Initializers/Normal/Normal.php	Add normal initializer with validation.
src/NeuralNet/Initializers/Normal/Exceptions/InvalidStandardDeviationException.php	Add custom exception for invalid std dev.
src/NeuralNet/Initializers/LeCun/LeCunUniform.php	Add LeCun uniform initializer.
src/NeuralNet/Initializers/LeCun/LeCunNormal.php	Add LeCun normal initializer.
src/NeuralNet/Initializers/He/HeUniform.php	Add He uniform initializer.
src/NeuralNet/Initializers/He/HeNormal.php	Add He normal initializer.
src/NeuralNet/Initializers/Constant/Constant.php	Add constant initializer.
src/NeuralNet/Initializers/Base/Initializer.php	Add Initializer interface.
src/NeuralNet/Initializers/Base/Exceptions/InvalidFanOutException.php	Add invalid fanOut exception.
src/NeuralNet/Initializers/Base/Exceptions/InvalidFanInException.php	Add invalid fanIn exception.
src/NeuralNet/Initializers/Base/AbstractInitializer.php	Add shared fanIn/fanOut validation base class.
src/NeuralNet/CostFunctions/RelativeEntropy/RelativeEntropy.php	Add KL-divergence style classification loss.
src/NeuralNet/CostFunctions/MeanAbsoluteError/MeanAbsoluteError.php	Add MAE regression loss.
src/NeuralNet/CostFunctions/LeastSquares/LeastSquares.php	Add least-squares regression loss.
src/NeuralNet/CostFunctions/HuberLoss/Exceptions/InvalidAlphaException.php	Add custom exception for invalid HuberLoss alpha.
src/NeuralNet/CostFunctions/Base/Contracts/RegressionLoss.php	Add RegressionLoss marker contract.
src/NeuralNet/CostFunctions/Base/Contracts/CostFunction.php	Add CostFunction interface contract.
src/NeuralNet/CostFunctions/Base/Contracts/ClassificationLoss.php	Add ClassificationLoss marker contract.
src/NeuralNet/ActivationFunctions/ThresholdedReLU/ThresholdedReLU.php	Add ThresholdedReLU activation function.
src/NeuralNet/ActivationFunctions/ThresholdedReLU/Exceptions/InvalidThresholdException.php	Add invalid threshold exception.
src/NeuralNet/ActivationFunctions/Softsign/Softsign.php	Add Softsign activation function.
src/NeuralNet/ActivationFunctions/Softplus/Softplus.php	Add Softplus activation function.
src/NeuralNet/ActivationFunctions/Sigmoid/Sigmoid.php	Add Sigmoid activation function (output-buffer derivative).
src/NeuralNet/ActivationFunctions/SiLU/SiLU.php	Add SiLU activation function.
src/NeuralNet/ActivationFunctions/ReLU6/ReLU6.php	Add ReLU6 activation function.
src/NeuralNet/ActivationFunctions/ReLU/ReLU.php	Add ReLU activation function.
src/NeuralNet/ActivationFunctions/LeakyReLU/Exceptions/InvalidLeakageException.php	Add invalid leakage exception.
src/NeuralNet/ActivationFunctions/HyperbolicTangent/HyperbolicTangent.php	Add tanh activation function.
src/NeuralNet/ActivationFunctions/HardSiLU/HardSiLU.php	Add HardSiLU activation function.
src/NeuralNet/ActivationFunctions/ELU/Exceptions/InvalidAlphaException.php	Add invalid alpha exception.
src/NeuralNet/ActivationFunctions/Base/Contracts/OBufferDerivative.php	Add output-buffer derivative contract.
src/NeuralNet/ActivationFunctions/Base/Contracts/IOBufferDerivative.php	Add input/output-buffer derivative contract.
src/NeuralNet/ActivationFunctions/Base/Contracts/IBufferDerivative.php	Add input-buffer derivative contract.
src/NeuralNet/ActivationFunctions/Base/Contracts/Derivative.php	Add derivative marker interface.
src/NeuralNet/ActivationFunctions/Base/Contracts/ActivationFunction.php	Add ActivationFunction interface contract.
src/Loggers/Screen.php	Update logger to accept Stringable messages safely.
src/Loggers/BlackHole.php	Update logger signature to accept Stringable messages.
src/Helpers/Stats.php	Tighten typing and use null-coalescing assignment.
src/Helpers/Params.php	Update toString signature to accept mixed.
src/GridSearch.php	Add __call() return type mixed.
src/Graph/Nodes/Split.php	Add union typing for split value and return type.
src/Graph/Nodes/Outcome.php	Add return type to outcome() in interface.
src/Graph/Nodes/Isolator.php	Add union typing for split value and return type.
src/Graph/Nodes/Depth.php	Remove stale/incorrect `@throws` annotation.
src/Graph/Nodes/Box.php	Add union typing for split value and return type.
src/Graph/Nodes/Best.php	Remove stale/incorrect `@throws` annotation.
src/Graph/Nodes/Average.php	Add union typing and return type.
src/Extractors/NDJSON.php	Add overwrite flag; switch to append-by-default.
src/Extractors/Exporter.php	Add overwrite parameter to Exporter interface.
src/Extractors/CSV.php	Add overwrite flag; switch to append-by-default; improve errors.
src/Exceptions/ClassRevisionMismatch.php	Improve exception message clarity.
src/Deferred.php	Add mixed return types for compute() and __invoke().
src/Datasets/Unlabeled.php	Type splitByFeature value parameter.
src/Datasets/Labeled.php	Add return type for label(); type splitByFeature value parameter.
src/Datasets/Dataset.php	Add overwrite flag to exportTo(); type splitByFeature value parameter.
src/DataType.php	Use mixed typing and adjust type detection logic.
src/CrossValidation/Metrics/Metric.php	Docblock tag ordering tweak.
src/CommitteeMachine.php	Add decideDiscrete return type.
src/Clusterers/MeanShift.php	Use null-coalescing assignment.
src/Classifiers/SoftmaxClassifier.php	Refactor network type to Network interface.
src/Classifiers/SVC.php	Use null-coalescing assignment.
src/Classifiers/LogisticRegression.php	Refactor network type to Network interface; add __unserialize workaround.
src/Classifiers/AdaBoost.php	Update phpdoc types; use nullsafe logger call.
src/BootstrapAggregator.php	Add typed constants/docs; improve typing around backend processing.
src/Backends/Tasks/TrainAndValidate.php	Clarify labels type and use named args for metric score.
src/Backends/Serial.php	Type backend enqueue context parameter as mixed.
src/Backends/Backend.php	Type backend enqueue context parameter as mixed.
src/Backends/Amp.php	Use null-coalescing assignment; type context parameters as mixed.
src/AnomalyDetectors/RobustZScore.php	Add typed constant; adjust epsilon smoothing code.
src/AnomalyDetectors/OneClassSVM.php	Use null-coalescing assignment.
src/AnomalyDetectors/Loda.php	Adjust phpdoc return array types and add inline non-empty annotation.
src/AnomalyDetectors/GaussianMLE.php	Adjust epsilon smoothing code.
phpunit.xml	Rework PHPUnit configuration (source/test suites/display details).
phpstan.neon	Add baseline, PHP version config, and exclude Swoole paths.
docs/transformers/tf-idf-transformer.md	Update TF-IDF parameter name and related guidance.
docs/transformers/regex-filter.md	Fix predefined patterns table ordering/format.
docs/regressors/mlp-regressor.md	Update MLP regressor params (eval interval / remove l2Penalty).
docs/regressors/gradient-boost.md	Update GBM params (eval interval).
docs/neural-network/optimizers/stochastic.md	Update optimizer namespace and add math section.
docs/neural-network/optimizers/step-decay.md	Update optimizer namespace and add math section.
docs/neural-network/optimizers/rms-prop.md	Update optimizer namespace and add math section.
docs/neural-network/optimizers/momentum.md	Update optimizer namespace and add math section.
docs/neural-network/optimizers/cyclical.md	Update optimizer namespace and add math section.
docs/neural-network/optimizers/adamax.md	Update optimizer namespace and add math section.
docs/neural-network/optimizers/adam.md	Update optimizer namespace and add math section.
docs/neural-network/optimizers/adagrad.md	Update optimizer namespace and add math section.
docs/neural-network/hidden-layers/swish.md	Update layer/initializer namespaces.
docs/neural-network/hidden-layers/prelu.md	Update layer/initializer namespaces.
docs/neural-network/hidden-layers/placeholder1d.md	Add Placeholder1D layer documentation.
docs/neural-network/hidden-layers/noise.md	Update layer namespace.
docs/neural-network/hidden-layers/dropout.md	Update layer namespace.
docs/neural-network/hidden-layers/dense.md	Update layer namespace.
docs/neural-network/hidden-layers/batch-norm.md	Update layer/initializer namespaces.
docs/neural-network/hidden-layers/activation.md	Update layer/activation namespaces.
docs/neural-network/cost-functions/relative-entropy.md	Update cost function namespace.
docs/neural-network/cost-functions/mean-absolute-error.md	Add MAE cost function documentation.
docs/neural-network/cost-functions/least-squares.md	Update cost function namespace.
docs/neural-network/cost-functions/huber-loss.md	Reformat math block for clarity.
docs/neural-network/cost-functions/cross-entropy.md	Update cost function namespace.
docs/neural-network/activation-functions/thresholded-relu.md	Update namespace, add plots, improve math.
docs/neural-network/activation-functions/softsign.md	Update namespace, add plots, references formatting.
docs/neural-network/activation-functions/softplus.md	Add Softplus documentation.
docs/neural-network/activation-functions/softmax.md	Update namespace, improve math/plots.
docs/neural-network/activation-functions/soft-plus.md	Remove legacy Soft Plus doc (replaced by softplus).
docs/neural-network/activation-functions/silu.md	Update namespace, add math/plots.
docs/neural-network/activation-functions/sigmoid.md	Update namespace, add plots.
docs/neural-network/activation-functions/selu.md	Update namespace, improve math/plots.
docs/neural-network/activation-functions/relu6.md	Add ReLU6 documentation.
docs/neural-network/activation-functions/relu.md	Update namespace, improve math/plots and refs.
docs/neural-network/activation-functions/leaky-relu.md	Update namespace, improve math/plots and refs.
docs/neural-network/activation-functions/hyperbolic-tangent.md	Update namespace, add plots.
docs/neural-network/activation-functions/hard-silu.md	Add HardSiLU documentation.
docs/neural-network/activation-functions/hard-sigmoid.md	Add HardSigmoid documentation.
docs/neural-network/activation-functions/gelu.md	Update namespace, add math/plots.
docs/neural-network/activation-functions/elu.md	Update namespace, improve math/plots and refs.
docs/installation.md	Update docs to PHP 8+ stated requirement.
docs/extractors/api.md	Document new overwrite behavior for exporters.
docs/exploring-data.md	Adjust dataset export examples for overwrite flag.
docs/datasets/api.md	Document dataset exportTo overwrite flag.
docs/classifiers/multilayer-perceptron.md	Update MLP classifier params (eval interval / remove l2Penalty).
docs/classifiers/logit-boost.md	Update LogitBoost params (eval interval).
docs/backends/swoole.md	Add backend documentation for Swoole.
composer.json	Bump PHP requirement, add ext-rubixnumpower, update dev tooling versions.
benchmarks/Transformers/TSNEBench.php	Fix phpdoc syntax.
benchmarks/Tokenizers/WordStemmerBench.php	Add typed constants/properties and imports.
benchmarks/Tokenizers/WordBench.php	Add typed constants/properties and imports.
benchmarks/Tokenizers/WhitespaceBench.php	Add typed constants/properties and imports.
benchmarks/Tokenizers/SentenceBench.php	Add typed constants/properties and imports.
benchmarks/Tokenizers/NGramBench.php	Add typed constants/properties and imports.
benchmarks/Tokenizers/KSkipNGramBench.php	Add typed constants/properties and imports.
benchmarks/Regressors/SVRBench.php	Add typed constants/properties and imports.
benchmarks/Regressors/RidgeBench.php	Add typed constants/properties and imports.
benchmarks/Regressors/RegressionTreeBench.php	Add typed constants/properties and imports.
benchmarks/Regressors/RadiusNeighborsRegressorBench.php	Add typed constants/properties and imports.
benchmarks/Regressors/MLPRegressorBench.php	Add typed constants/properties and imports.
benchmarks/Regressors/KNNRegressorBench.php	Add typed constants/properties and imports.
benchmarks/Regressors/KDNeighborsRegressorBench.php	Add typed constants/properties and imports.
benchmarks/Regressors/GradientBoostBench.php	Add typed constants/properties and imports.
benchmarks/Regressors/ExtraTreeRegressorBench.php	Add typed constants/properties and imports.
benchmarks/Regressors/AdalineBench.php	Add typed constants/properties and imports.
benchmarks/Graph/Trees/VantageTreeBench.php	Fix phpdoc syntax.
benchmarks/Graph/Trees/KDTreeBench.php	Fix phpdoc syntax.
benchmarks/Graph/Trees/BallTreeBench.php	Fix phpdoc syntax.
benchmarks/Datasets/SplittingBench.php	Fix phpdoc syntax.
benchmarks/Datasets/SortingBench.php	Fix phpdoc syntax.
benchmarks/Datasets/RandomizationBench.php	Fix phpdoc syntax.
benchmarks/Clusterers/MeanShiftBench.php	Add typed constants/properties and imports.
benchmarks/Classifiers/SoftmaxClassifierBench.php	Add typed constants/properties and imports.
benchmarks/Classifiers/SVCBench.php	Add typed constants/properties and imports.
benchmarks/Classifiers/RandomForestBench.php	Add backend param provider integration; typed properties.
benchmarks/Classifiers/RadiusNeighborsBench.php	Add typed constants/properties and imports.
benchmarks/Classifiers/OneVsRestBench.php	Add backend param provider integration; typed properties.
benchmarks/Classifiers/NaiveBayesBench.php	Add typed constants/properties and imports.
benchmarks/Classifiers/MultilayerPerceptronBench.php	Add typed constants/properties and imports.
benchmarks/Classifiers/LogitBoostBench.php	Add typed constants/properties and imports.
benchmarks/Classifiers/LogisticRegressionBench.php	Add typed constants/properties and imports.
benchmarks/Classifiers/KNearestNeighborsBench.php	Add typed constants/properties and imports.
benchmarks/Classifiers/KDNeighborsBench.php	Add typed constants/properties and imports.
benchmarks/Classifiers/GaussianNBBench.php	Add typed constants/properties and imports.
benchmarks/Classifiers/ExtraTreeClassifierBench.php	Add typed constants/properties and imports.
benchmarks/Classifiers/ClassificationTreeBench.php	Add typed constants/properties and imports.
benchmarks/Classifiers/AdaBoostBench.php	Add typed constants/properties and imports.
benchmarks/AnomalyDetectors/RobustZScoreBench.php	Add typed constants/properties and imports.
benchmarks/AnomalyDetectors/OneClassSVMBench.php	Add typed constants/properties and imports.
benchmarks/AnomalyDetectors/LodaBench.php	Add typed constants/properties and imports.
benchmarks/AnomalyDetectors/LocalOutlierFactorBench.php	Add typed constants/properties and imports.
benchmarks/AnomalyDetectors/IsolationForestBench.php	Add typed constants/properties and imports.
benchmarks/AnomalyDetectors/GaussianMLEBench.php	Add typed constants/properties and imports.
README.md	Update stated PHP requirement.
CHANGELOG.md	Add 3.0.0 release notes and other entries.
.phplint.yml	Add PHPLint configuration.
.php-cs-fixer.dist.php	Update CS fixer finder/rules for PHP 8.4 and caching.
.github/workflows/ci.yml	Update CI matrix and add NumPower build/install steps.

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

Copilot · 2026-03-10T23:45:10Z

src/functions.php

+    function iterator_first(iterable $iterator) : mixed
    {
        foreach ($iterator as $element) {
            return $element;
        }
+
+        throw new RuntimeException('Iterator did not return any elements.');
    }


In a namespaced file, RuntimeException resolves to Rubix\\ML\\RuntimeException unless qualified/imported. This will fatally error if that class does not exist. Use \\RuntimeException (global) or import/qualify the intended exception class.

Copilot · 2026-03-10T23:45:11Z

src/Regressors/Ridge.php

     */
    protected ?Vector $coefficients = null;

+    protected ?nd $coefficientsNd = null;


$this->coefficientsNd appears to include the bias term (since you array_shift() the first element into $this->bias but keep $this->coefficientsNd unchanged). In predict(), you dot datasetNd (shape: n_samples × n_features) with $this->coefficientsNd (likely length n_features+1), which will cause a shape mismatch or incorrect results. Fix by either (a) storing $this->coefficientsNd without the bias term, or (b) augmenting the dataset matrix with a bias column (ones) and omitting the explicit + $this->bias.

Copilot · 2026-03-10T23:45:11Z

src/Regressors/Ridge.php


        $x = Matrix::build($dataset->samples())->augmentLeft($biases);
-        $y = Vector::build($dataset->labels());
+        $y = nd::array($dataset->labels());


$this->coefficientsNd appears to include the bias term (since you array_shift() the first element into $this->bias but keep $this->coefficientsNd unchanged). In predict(), you dot datasetNd (shape: n_samples × n_features) with $this->coefficientsNd (likely length n_features+1), which will cause a shape mismatch or incorrect results. Fix by either (a) storing $this->coefficientsNd without the bias term, or (b) augmenting the dataset matrix with a bias column (ones) and omitting the explicit + $this->bias.

Copilot · 2026-03-10T23:45:11Z

src/Regressors/Ridge.php

+        $this->coefficientsNd = nd::dot($xMulAddInv, $xtDotY);
+        $coefficients = $this->coefficientsNd->toArray();

        $this->bias = (float) array_shift($coefficients);


$this->coefficientsNd appears to include the bias term (since you array_shift() the first element into $this->bias but keep $this->coefficientsNd unchanged). In predict(), you dot datasetNd (shape: n_samples × n_features) with $this->coefficientsNd (likely length n_features+1), which will cause a shape mismatch or incorrect results. Fix by either (a) storing $this->coefficientsNd without the bias term, or (b) augmenting the dataset matrix with a bias column (ones) and omitting the explicit + $this->bias.

Suggested change

$this->bias = (float) array_shift($coefficients);

$this->bias = (float) array_shift($coefficients);

$this->coefficientsNd = nd::array($coefficients);

Copilot · 2026-03-10T23:45:11Z

src/Regressors/Ridge.php

+        $datasetNd = nd::array($dataset->samples());
+        $datasetDotCoefficients = nd::dot($datasetNd, $this->coefficientsNd);
+
+        return nd::add($datasetDotCoefficients, $this->bias)->toArray();


$this->coefficientsNd appears to include the bias term (since you array_shift() the first element into $this->bias but keep $this->coefficientsNd unchanged). In predict(), you dot datasetNd (shape: n_samples × n_features) with $this->coefficientsNd (likely length n_features+1), which will cause a shape mismatch or incorrect results. Fix by either (a) storing $this->coefficientsNd without the bias term, or (b) augmenting the dataset matrix with a bias column (ones) and omitting the explicit + $this->bias.

Copilot · 2026-03-10T23:45:15Z

src/DataType.php

-
-                return new self(self::OTHER);
-
            default:


Image detection via GD appears to have been removed entirely. In PHP 8+, GD commonly returns a \\GdImage object (not a resource), so falling through to OTHER will regress image support. Consider explicitly handling object values that are instances of \\GdImage to return self::IMAGE (or equivalent).

Copilot · 2026-03-10T23:45:15Z

src/DataType.php

                return new self(self::OTHER);
        }


Image detection via GD appears to have been removed entirely. In PHP 8+, GD commonly returns a \\GdImage object (not a resource), so falling through to OTHER will regress image support. Consider explicitly handling object values that are instances of \\GdImage to return self::IMAGE (or equivalent).

Copilot · 2026-03-10T23:45:15Z

CHANGELOG.md

+    - Remove Network interface
+    - RBX Serializer only tracks major library version number
+    - Convert NeuralNet classes to use NDArray instead of Matrix
+    - Turn back Network interface


These entries are contradictory as written and make it unclear what actually changed in 3.0.0. Please reconcile into a single accurate bullet (e.g., 'Replaced concrete Network with interface', or 'Renamed/moved Network contract') to avoid confusing release notes.

Suggested change

- Turn back Network interface

Copilot · 2026-03-10T23:45:15Z

src/Loggers/BlackHole.php

@@ -20,7 +22,7 @@ class BlackHole extends Logger
     * @param string $message


The docblock still says @param string $message but the method signature now accepts string|Stringable. Update the docblock to match the new accepted types.

Suggested change

* @param string $message

* @param string|Stringable $message

Copilot · 2026-03-10T23:45:16Z

tests/Backends/SwooleTest.php

+use Rubix\ML\Specifications\SwooleExtensionIsLoaded;
+use Swoole\Event;
+
+#[Group('backends')]


Group naming is inconsistent with other backend tests using #[Group('Backends')] (capitalized). Consider standardizing group names/casing to make filtered runs (--group) predictable.

Suggested change

#[Group('backends')]

#[Group('Backends')]

andrewdalpino and others added 30 commits May 27, 2023 15:54

Only track major version number

e0957f1

Fix branch

04d77b4

Merge branch 'master' into 3.0

c014c91

Initial commit (RubixML#299)

665a3d1

Merge branch '2.5' of github.com:/RubixML/ML into 2.5

a14e45d

Vantage Tree (RubixML#300)

92bbebe

* Initial commit * Better testing * Improve the docs * Rename benchmark * Explicitly import max() function

Merge branch 'master' into 2.5

685b53d

Merge branch '2.5' into 3.0

bb6ef64

Merge branch 'master' into 2.5

89b6051

Merge branch '2.5' into 3.0

4f9fb9d

Merge branch 'master' into 2.5

20aba67

Merge branch '2.5' into 3.0

1381f5c

Fix coding style

865e54a

Merge branch '2.5' into 3.0

57e1811

Wrapper interface (RubixML#314)

4c16268

* Add Wrapper interface for models wrappers * Add WrapperAware trait * Fix PhpDoc * Revert "Add WrapperAware trait" This reverts commit 241abc4. * Rename Wrapper interface to EstimatorWrapper * PHP CS fix

Plus plus check (RubixML#317)

277e37b

* Initial commit * Allow deltas in units tests

Merge branch 'master' into 2.5

e3790bd

Merge branch '2.5' of https://github.com/RubixML/ML into 2.5

0fa98ce

Merge branch '2.5' into 3.0

dd56e81

Fix coding style and composer.lock

bf5d32c

fix(swoole): setAffinity does not exist on some versions of Swoole (R…

274076c

…ubixML#327)

Merge branch 'master' into 2.5

8e299a2

Merge branch '2.5' into 3.0

4033a9d

Fix tests and such

19aecd1

Merge branch 'master' into 3.0

15726a4

Add numpower to the GitHub workflow pipeline (RubixML#331)

37730cb

* chore: add NumPower to workflow pipeline. * chore: install openbla. * chore: install openblas from apt repo.

Allow normalizers to skip NaN values (RubixML#333)

646b1a2

* Add is_finite checks to skip NAN values * Tests and fix issue with samples that are all non-finite

chore: use NumPower (RubixML#315)

ca59648

apphp added 23 commits January 25, 2026 16:22

ML-394 Converted FeedForward to use NumPower

106c070

ML-394 Converted Network to use NumPower

e118309

ML-394 Converted Snapshot to use NumPower

5dfc354

ML-394 Improved FeedForwardTest

0e815da

ML-394 Fixed copilot style comments, added prevention division by zer…

a261d69

…o in Swish

ML-396 Converted MLPRegressor to work with NumPower/NDArray related c…

c80d278

…lasses

ML-396 removed unneeded export function

13acae6

ML-396 added test for NumPower

3b65a47

ML-396 added USE_NUMPOWER_TRANSPOSE option to Network

d7404f8

ML-396 added USE_NUMPOWER_TRANSPOSE option to Network

d538799

ML-396 fixed issue with samples normalization

f333c67

ML-396 removed unneeded packages from composer

1583ee3

ML-396 removed unneeded packages from composer

57037c6

ML-392 rewritten BatchNorm::gradient for better readability

7fedecd

ML-392 rewritten BatchNorm::forward for better readability

ee43ccc

ML-392 rewritten BatchNorm to use the library-wide convention [num_fe…

a8a1a54

…atures, num_samples]

ML-392 added comment for optimization specific to the Binary Cross En…

0ce683c

…tropy loss function

Merge pull request RubixML#393 from apphp/392-convert-layers-to-numpower

318efc4

392 convert layers to numpower

Merge branch '3.0' into 394-convert-misc-nn-classes-to-NumPower

97c2173

Merge branch '3.0' into 396-convert-mlp-classifier-to-NumPower

22df3e0

Merge pull request RubixML#395 from apphp/394-convert-misc-nn-classes…

f59be22

…-to-NumPower 394 convert misc nn classes to num power

Merge branch '3.0' into 396-convert-mlp-classifier-to-NumPower

b79ddb6

ML-400 convert Network to interface

66ce8ec

apphp requested review from andrewdalpino and Copilot March 10, 2026 23:41

apphp self-assigned this Mar 10, 2026

apphp closed this Mar 10, 2026

apphp reopened this Mar 10, 2026

apphp closed this Mar 10, 2026

Copilot AI reviewed Mar 10, 2026

View reviewed changes

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

400 refactor network class to interface#401

400 refactor network class to interface#401
apphp wants to merge 302 commits intoRubixML:masterfrom
apphp:400-refactor-network-class-to-interface

apphp commented Mar 10, 2026

Uh oh!

Copilot AI left a comment

Uh oh!

Copilot AI Mar 10, 2026

Uh oh!

Copilot AI Mar 10, 2026

Uh oh!

Copilot AI Mar 10, 2026

Uh oh!

Copilot AI Mar 10, 2026

Uh oh!

Copilot AI Mar 10, 2026

Uh oh!

Copilot AI Mar 10, 2026

Uh oh!

Copilot AI Mar 10, 2026

Uh oh!

Copilot AI Mar 10, 2026

Uh oh!

Copilot AI Mar 10, 2026

Uh oh!

Copilot AI Mar 10, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

8 participants

	$this->bias = (float) array_shift($coefficients);
	$this->bias = (float) array_shift($coefficients);
	$this->coefficientsNd = nd::array($coefficients);

		@@ -20,7 +22,7 @@ class BlackHole extends Logger
		* @param string $message

Uh oh!

Conversation

apphp commented Mar 10, 2026

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Reviewed changes

Uh oh!

Copilot AI Mar 10, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Mar 10, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Mar 10, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Mar 10, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Mar 10, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Mar 10, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Mar 10, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Mar 10, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Mar 10, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Mar 10, 2026

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

8 participants