Prepare for 0.1.0-beta release

CHANGELOG.md

@@ -1,4 +1,10 @@
-- Unreleased
+- 0.1.0-beta
+    - Add Recursive Feature Eliminator feature selector
     - Implement BM25 TF-IDF Transformer
+    - Add Delta TF-IDF Transformer
     - Added Lambda function Transformer
-    - All objects implement Stringable interface
+    - All objects implement Stringable interface
+    - Added Gower nan-safe distance kernel
+    - Added ISRU and ISRLU activation functions
+    - Added Alpha Dropout hidden layer
+    - Added Model Orchestra meta-estimator

composer.json

@@ -16,7 +16,7 @@
     ],
     "require": {
         "php": ">=7.2",
-        "rubix/ml": "^0.1.0-rc3",
+        "rubix/ml": "^0.1.0",
         "rubix/tensor": "^2.0.4",
         "wamania/php-stemmer": "^2.0"
     },

docs/kernels/distance/gower.md

@@ -16,11 +16,7 @@ A robust distance kernel that measures a mix of categorical and continuous data
 ```php
 use Rubix\ML\Kernels\Distance\Gower;
 
-$kernel = new Gower(); // Continuous features between 0 and 1
-
-$kernel = new Gower(2.0); // Between -1 and 1
-
-$kernel = new Gower(1000.0); // Between 0 and 1000
+$kernel = new Gower(2.0);
 ```
 
 ### References

docs/transformers/bm25-transformer.md

@@ -1,4 +1,4 @@
-<span style="float:right;"><a href="https://github.com/RubixML/RubixML/blob/master/src/Transformers/TfIdfTransformer.php">[source]</a></span>
+<span style="float:right;"><a href="https://github.com/RubixML/Extras/blob/master/src/Transformers/BM25Transformer.php">[source]</a></span>
 
 # BM25 Transformer
 BM25 is a term frequency weighting scheme that takes term frequency (TF) saturation and document length into account.

docs/transformers/delta-tf-idf-transformer.md

@@ -0,0 +1,30 @@
+<span style="float:right;"><a href="https://github.com/RubixML/Extras/blob/master/src/Transformers/DeltaTfIdfTransformer.php">[source]</a></span>
+
+# Delta TF-IDF Transformer
+A supervised TF-IDF (Term Frequency Inverse Document Frequency) Transformer that uses class labels to boost the TF-IDFs of terms by how informative they are. Terms that receive the highest boost are those whose concentration is primarily in one class whereas low weighted terms are more evenly distributed among the classes.
+
+> **Note:** This transformer assumes that its input is made up of word frequency vectors such as those produced by [Word Count Vectorizer](word-count-vectorizer.md).
+
+**Interfaces:** [Transformer](api.md#transformer), [Stateful](api.md#stateful), [Elastic](api.md#elastic)
+
+**Data Type Compatibility:** Continuous only
+
+## Parameters
+This transformer does not have any parameters.
+
+## Example
+```php
+use Rubix\ML\Transformers\DeltaTfIdfTransformer;
+
+$transformer = new DeltaTfIdfTransformer();
+```
+
+## Additional Methods
+Return the document frequencies calculated during fitting:
+```php
+public dfs() : ?array
+```
+
+### References
+>- J. Martineau et al. (2009). Delta TFIDF: An Improved Feature Space for Sentiment Analysis.
+>- S. Ghosh et al. (2018). Class Specific TF-IDF Boosting for Short-text Classification.

docs/transformers/lambda-function.md

@@ -1,4 +1,4 @@
-<span style="float:right;"><a href="https://github.com/RubixML/RubixML/blob/master/src/Transformers/LambdaFunction.php">[source]</a></span>
+<span style="float:right;"><a href="https://github.com/RubixML/Extras/blob/master/src/Transformers/LambdaFunction.php">[source]</a></span>
 
 # Lambda Function
 Run a stateless lambda function (*anonymous* function) over the samples. The lambda function receives the sample matrix as an argument and should return the transformed matrix.

src/Kernels/Distance/Gower.php

@@ -117,6 +117,6 @@ public function compute(array $a, array $b) : float
      */
     public function __toString() : string
     {
-        return "Gower {range: {$this->range}}";
+        return "Gower (range: {$this->range})";
     }
 }

src/ModelOrchestra.php

@@ -384,6 +384,6 @@ protected function extractRegressor(Dataset $dataset) : array
      */
     public function __toString() : string
     {
-        return 'Model Orchestra {' . Params::stringify($this->params()) . '}';
+        return 'Model Orchestra (' . Params::stringify($this->params()) . ')';
     }
 }

src/NeuralNet/ActivationFunctions/ISRLU.php

@@ -95,6 +95,6 @@ public function differentiate(Matrix $z, Matrix $computed) : Matrix
      */
     public function __toString() : string
     {
-        return "ISRLU {alpha: {$this->alpha}}";
+        return "ISRLU (alpha: {$this->alpha})";
     }
 }

src/NeuralNet/ActivationFunctions/ISRU.php

@@ -95,6 +95,6 @@ public function differentiate(Matrix $z, Matrix $computed) : Matrix
      */
     public function __toString() : string
     {
-        return "ISRU {alpha: {$this->alpha}}";
+        return "ISRU (alpha: {$this->alpha})";
     }
 }

src/NeuralNet/Layers/AlphaDropout.php

@@ -200,6 +200,6 @@ public function saturate(int $value) : float
      */
     public function __toString() : string
     {
-        return "Alpha Dropout {ratio: {$this->ratio}}";
+        return "Alpha Dropout (ratio: {$this->ratio})";
     }
 }

src/Transformers/DeltaTfIdfTransformer.php

@@ -32,7 +32,7 @@
  * @package     Rubix/ML
  * @author      Andrew DalPino
  */
-class DeltaTfIdfTransformer implements Elastic
+class DeltaTfIdfTransformer implements Transformer, Stateful, Elastic
 {
     /**
      * The class specific term frequencies of each word i.e. the number of
@@ -102,6 +102,16 @@ public function fitted() : bool
         return $this->idfs and $this->entropies;
     }
 
+    /**
+     * Return the document frequencies calculated during fitting.
+     *
+     * @return int[]|null
+     */
+    public function dfs() : ?array
+    {
+        return $this->dfs;
+    }
+
     /**
      * Fit the transformer to the dataset.
      *