396 convert regressors to num power by apphp · Pull Request #411 · RubixML/ML

apphp · 2026-03-28T21:22:27Z

No description provided.

…lasses

# Conflicts: # src/NeuralNet/Networks/FeedForward/FeedForward.php # tests/NeuralNet/Layers/Swish/SwishTest.php # tests/NeuralNet/Networks/NetworkTest.php

andrewdalpino · 2026-07-06T19:58:40Z

@@ -0,0 +1,58 @@
+<?php
+
+namespace Rubix\ML\Benchmarks\NeuralNet\ActivationFunctions\ELU;


Let's change the path of this file to reflect it's namespace i.e. remove the ELU folder.

andrewdalpino · 2026-07-06T19:59:13Z

@@ -0,0 +1,58 @@
+<?php
+
+namespace Rubix\ML\Benchmarks\NeuralNet\ActivationFunctions\GELU;


The same comment to ELU applies to each of these benchmark files.

andrewdalpino · 2026-07-06T20:00:28Z

     * @var Matrix
     */
-    protected $z;
+    protected Matrix $z;


Since we are swapping Tensor out for NumPower, we won't need these old benchmarks.

andrewdalpino · 2026-07-06T20:01:50Z

+        // New samples start with provisional label 0 in this partial batch,
+        // so size bookkeeping must include them before any reassignments.
+        $this->sizes[0] += $dataset->numSamples();
+


Why are we making edits to K Means as part of this?

andrewdalpino · 2026-07-06T20:05:59Z

+     *
+     * Cloning an NDArray directly may trigger native memory corruption in some
+     * NumPower builds (e.g. heap corruption/segfaults when parameters are
+     * snapshotted during training). To make cloning deterministic and stable we


Nice comment!

andrewdalpino · 2026-07-06T20:06:55Z

        $this->network = new FeedForward(
            new Placeholder1D($dataset->numFeatures()),
-            [new Dense(1, $this->l2Penalty, true, new Xavier2())],
+            [new Dense(1, $this->l2Penalty, true, new XavierUniform())],


Why are we renaming this class? Will it cause a BC break?

andrewdalpino · 2026-07-06T20:08:48Z

        ?RegressionLoss $costFn = null,
-        ?Metric $metric = null
+        ?Metric $metric = null,
+        bool $packSamples = false


Why are we adding this as a constructor argument? Can we autodetect if we need to pack the array instead?

andrewdalpino · 2026-07-06T20:09:43Z

-            foreach ($distances as $distance) {
-                $weights[] = 1.0 / (1.0 + $distance);
-            }
+            $distances = NumPower::array($distances);


Let's not convert this file to NumPower. It will not benefit as much as others and requires an extra dependency.

andrewdalpino · 2026-07-06T20:11:39Z

 * @package     Rubix/ML
 * @author      Andrew DalPino
+ * @author      Samuel Akopyan <leumas.a@gmail.com>
 */


Let's remove all these @author annotations (including mine). They will become meaningless as time goes on and we can use Git blame if necessary. If the purpose is for recognition, there are probably better ways we can go about it.

Actually on second thought, let's do that in a follow up commit. All good!

andrewdalpino · 2026-07-06T20:12:30Z

        -
            message: '#^Property Rubix\\ML\\Classifiers\\NaiveBayes\:\:\$counts \(array<string, list<array<int<0, max>>>>\) does not accept non\-empty\-array<array<array<int<0, max>>>>\.$#'
            identifier: assign.propertyType
+            count: 1


What is the reason for not using the same phpstan configuration for both local dev and CI?

andrewdalpino · 2026-07-06T20:14:47Z

+    'float64',
+    'int8',
+    'int16',
+    'int32',


Why do we need these constants?

andrewdalpino · 2026-07-06T20:15:18Z

@@ -19,8 +19,8 @@ parameters:
 			path: src/NeuralNet/Networks/FeedForward/FeedForward.php

 		-


What's the reason for this config?

andrewdalpino · 2026-07-06T20:16:24Z

@@ -11,6 +11,10 @@
 use Rubix\ML\Exceptions\RuntimeException;
 use PHPUnit\Framework\TestCase;


Why are we changing this file as part of this PR?

andrewdalpino · 2026-07-06T20:16:48Z

@@ -6,18 +6,20 @@

 use PHPUnit\Framework\Attributes\CoversClass;
 use PHPUnit\Framework\Attributes\Group;


Why are we changing this file as part of this PR?

andrewdalpino

Hey @apphp, it looks like there are alot of changes here that don't really have much to do with converting the Regressors to NumPower. Let's revert any changes that aren't necessary to complete the task at hand and we can follower up with other changes in separate PRs.

# Conflicts: # benchmarks/NeuralNet/ActivationFunctions/ELUBench.php # benchmarks/NeuralNet/ActivationFunctions/GELUBench.php # phpstan-baseline.neon # phpstan-ci.neon # src/NeuralNet/ActivationFunctions/Softmax/Softmax.php # src/NeuralNet/Initializers/Normal/Normal.php # src/NeuralNet/Initializers/Uniform/Uniform.php # src/NeuralNet/Layers/Noise/Noise.php # src/NeuralNet/Networks/FeedForward/FeedForward.php # src/NeuralNet/Parameters/Parameter.php # tests/NeuralNet/CostFunctions/CrossEntropy/CrossEntropyTest.php # tests/NeuralNet/CostFunctions/LeastSquares/LeastSquaresTest.php # tests/NeuralNet/Initializers/HeUniformTest.php # tests/NeuralNet/Initializers/Normal/NormalTest.php # tests/NeuralNet/Initializers/TruncatedNormalTest.php # tests/NeuralNet/Initializers/XavierNormalTest.php # tests/NeuralNet/Initializers/XavierUniformTest.php

apphp added 20 commits February 14, 2026 17:16

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

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

22df3e0

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

b79ddb6

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

cab6925

# Conflicts: # src/NeuralNet/Networks/FeedForward/FeedForward.php # tests/NeuralNet/Layers/Swish/SwishTest.php # tests/NeuralNet/Networks/NetworkTest.php

ML-396 style fixes

b920665

ML-396 migrated MLPRegressor

b99d65b

ML-396 migrated MLPRegressor

e04867b

ML-396 migrated MLPRegressor

36a282e

ML-396 migrated MLPRegressor

9bc5107

ML-396 migrated MLPRegressor

2a69709

ML-396 migrated Adaline

76093fd

ML-396 migrated Adaline

0d35e60

ML-396 migrated Adaline

289b822

apphp self-assigned this Mar 28, 2026

apphp added 9 commits March 29, 2026 23:40

ML-396 migrated ExtraTreeRegressor with Hyperplane

4e19926

ML-396 migrated RegressionTree

f18ceaa

ML-396 migrated GradientBoost

f22c58c

ML-396 migrated Ridge

8a24b57

ML-396 migrated Ridge

5052675

ML-396 migrated Ridge

4c31a38

ML-396 Added tests for Ridge

b6f3665

ML-396 Added AdalineTest

41b144c

ML-396 fixes for STAN

7017b0f

apphp added 11 commits June 24, 2026 23:09

ML-396 replaced old classes with new for SwissRoll

8acf3cd

ML-396 stabilized SparseRandomProjectorTest and GridSearchTest

a0194e1

ML-396 stabilized SparseRandomProjectorTest and GridSearchTest

394d4b9

ML-396 stabilized RadiusNeighborsTest and DBSCANTest

037d7ae

ML-396 stabilized KMeans and DBSCANTest

ce12de9

ML-396 stabilized VantageTreeTest

61116c5

ML-396 stabilized TruncatedNormalTest and BallTreeTest

0624dd8

ML-396 stabilized TruncatedNormalTest and VantageTreeTest

128f31f

ML-396 stabilized GaussianMixtureTest and RidgeProvider

a1d5330

ML-396 stabilized GaussianMixtureTest and RidgeProvider

97a3348

ML-396 stabilized GaussianMixtureTest and RidgeProvider

b58fea1

andrewdalpino reviewed Jul 6, 2026

View reviewed changes

Comment thread phpstan-bootstrap.php

'float64',

'int8',

'int16',

'int32',

andrewdalpino Jul 6, 2026

Copy link
Copy Markdown

Member

Choose a reason for hiding this comment

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

Why do we need these constants?

andrewdalpino reviewed Jul 6, 2026

View reviewed changes

andrewdalpino requested changes Jul 6, 2026

View reviewed changes

apphp added 3 commits July 7, 2026 01:06

ML-396 fixed class path, missing imports after merge with branch 3.0

b0347c5

ML-396 fixed class path, missing imports after merge with branch 3.0

d0804d4

apphp closed this Jul 7, 2026

		@@ -0,0 +1,58 @@
		<?php

		namespace Rubix\ML\Benchmarks\NeuralNet\ActivationFunctions\ELU;

		@@ -19,8 +19,8 @@ parameters:
		path: src/NeuralNet/Networks/FeedForward/FeedForward.php

		-

		@@ -11,6 +11,10 @@
		use Rubix\ML\Exceptions\RuntimeException;
		use PHPUnit\Framework\TestCase;

		@@ -6,18 +6,20 @@

		use PHPUnit\Framework\Attributes\CoversClass;
		use PHPUnit\Framework\Attributes\Group;

Uh oh!

Uh oh!

Conversation

apphp commented Mar 28, 2026

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

andrewdalpino Jul 6, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

andrewdalpino Jul 6, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

andrewdalpino left a comment

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

andrewdalpino Jul 6, 2026 •

edited

Loading

andrewdalpino Jul 6, 2026 •

edited

Loading