mlpack 4.6.1__cp38-cp38-win_amd64.whl → 4.6.2__cp38-cp38-win_amd64.whl

mlpack/include/mlpack/methods/ann/rnn.hpp

@@ -37,6 +37,7 @@ template<
 class RNN
 {
  public:
+  using CubeType = typename GetCubeType<MatType>::type;
   /**
    * Create the RNN object.
    *
@@ -128,8 +129,8 @@ class RNN
    */
   template<typename OptimizerType, typename... CallbackTypes>
   typename MatType::elem_type Train(
-      arma::Cube<typename MatType::elem_type> predictors,
-      arma::Cube<typename MatType::elem_type> responses,
+      CubeType predictors,
+      CubeType responses,
       OptimizerType& optimizer,
       CallbackTypes&&... callbacks);
 
@@ -156,8 +157,8 @@ class RNN
    */
   template<typename OptimizerType = ens::RMSProp, typename... CallbackTypes>
   typename MatType::elem_type Train(
-      arma::Cube<typename MatType::elem_type> predictors,
-      arma::Cube<typename MatType::elem_type> responses,
+      CubeType predictors,
+      CubeType responses,
       CallbackTypes&&... callbacks);
 
   /**
@@ -186,8 +187,8 @@ class RNN
    */
   template<typename OptimizerType, typename... CallbackTypes>
   typename MatType::elem_type Train(
-      arma::Cube<typename MatType::elem_type> predictors,
-      arma::Cube<typename MatType::elem_type> responses,
+      CubeType predictors,
+      CubeType responses,
       arma::urowvec sequenceLengths,
       OptimizerType& optimizer,
       CallbackTypes&&... callbacks);
@@ -222,8 +223,8 @@ class RNN
    */
   template<typename OptimizerType = ens::RMSProp, typename... CallbackTypes>
   typename MatType::elem_type Train(
-      arma::Cube<typename MatType::elem_type> predictors,
-      arma::Cube<typename MatType::elem_type> responses,
+      CubeType predictors,
+      CubeType responses,
       arma::urowvec sequenceLengths,
       CallbackTypes&&... callbacks);
 
@@ -236,8 +237,8 @@ class RNN
    * @param results Matrix to put output predictions of responses into.
    * @param batchSize Batch size to use for prediction.
    */
-  void Predict(const arma::Cube<typename MatType::elem_type>& predictors,
-               arma::Cube<typename MatType::elem_type>& results,
+  void Predict(const CubeType& predictors,
+               CubeType& results,
                const size_t batchSize = 128);
 
   /**
@@ -254,8 +255,8 @@ class RNN
    * @param predictors Input predictors.
    * @param results Matrix to put output predictions of responses into.
    */
-  void Predict(const arma::Cube<typename MatType::elem_type>& predictors,
-               arma::Cube<typename MatType::elem_type>& results,
+  void Predict(const CubeType& predictors,
+               CubeType& results,
                const arma::urowvec& sequenceLengths);
 
   // Return the nujmber of weights in the model.
@@ -318,8 +319,8 @@ class RNN
    * @param responses Target outputs for input variables.
    */
   typename MatType::elem_type Evaluate(
-      const arma::Cube<typename MatType::elem_type>& predictors,
-      const arma::Cube<typename MatType::elem_type>& responses);
+      const CubeType& predictors,
+      const CubeType& responses);
 
   //! Serialize the model.
   template<typename Archive>
@@ -425,8 +426,8 @@ class RNN
    * @param sequenceLengths (Optional) sequence length for each predictor
    *     sequence.
    */
-  void ResetData(arma::Cube<typename MatType::elem_type> predictors,
-                 arma::Cube<typename MatType::elem_type> responses,
+  void ResetData(CubeType predictors,
+                 CubeType responses,
                  arma::urowvec sequenceLengths = arma::urowvec());
 
  private:
@@ -456,11 +457,11 @@ class RNN
   // The matrix of data points (predictors).  These members are empty, except
   // during training---we must store a local copy of the training data since
   // the ensmallen optimizer will not provide training data.
-  arma::Cube<typename MatType::elem_type> predictors;
+  CubeType predictors;
 
   // The matrix of responses to the input data points.  This member is empty,
   // except during training.
-  arma::Cube<typename MatType::elem_type> responses;
+  CubeType responses;
 
   // The length of each input sequence.  If this is empty, then every sequence
   // is assuemd to have the same length (`predictors.n_slices`).

mlpack/include/mlpack/methods/ann/rnn_impl.hpp

@@ -156,8 +156,8 @@ typename MatType::elem_type RNN<
     InitializationRuleType,
     MatType
 >::Train(
-    arma::Cube<typename MatType::elem_type> predictors,
-    arma::Cube<typename MatType::elem_type> responses,
+    CubeType predictors,
+    CubeType responses,
     OptimizerType& optimizer,
     CallbackTypes&&... callbacks)
 {
@@ -190,8 +190,8 @@ typename MatType::elem_type RNN<
     InitializationRuleType,
     MatType
 >::Train(
-    arma::Cube<typename MatType::elem_type> predictors,
-    arma::Cube<typename MatType::elem_type> responses,
+    CubeType predictors,
+    CubeType responses,
     CallbackTypes&&... callbacks)
 {
   OptimizerType optimizer;
@@ -210,8 +210,8 @@ typename MatType::elem_type RNN<
     InitializationRuleType,
     MatType
 >::Train(
-    arma::Cube<typename MatType::elem_type> predictors,
-    arma::Cube<typename MatType::elem_type> responses,
+    CubeType predictors,
+    CubeType responses,
     arma::urowvec sequenceLengths,
     OptimizerType& optimizer,
     CallbackTypes&&... callbacks)
@@ -246,8 +246,8 @@ typename MatType::elem_type RNN<
     InitializationRuleType,
     MatType
 >::Train(
-    arma::Cube<typename MatType::elem_type> predictors,
-    arma::Cube<typename MatType::elem_type> responses,
+    CubeType predictors,
+    CubeType responses,
     arma::urowvec sequenceLengths,
     CallbackTypes&&... callbacks)
 {
@@ -266,8 +266,8 @@ void RNN<
     InitializationRuleType,
     MatType
 >::Predict(
-    const arma::Cube<typename MatType::elem_type>& predictors,
-    arma::Cube<typename MatType::elem_type>& results,
+    const CubeType& predictors,
+    CubeType& results,
     const size_t batchSize)
 {
   // Ensure that the network is configured correctly.
@@ -313,8 +313,8 @@ void RNN<
     InitializationRuleType,
     MatType
 >::Predict(
-    const arma::Cube<typename MatType::elem_type>& predictors,
-    arma::Cube<typename MatType::elem_type>& results,
+    const CubeType& predictors,
+    CubeType& results,
     const arma::urowvec& sequenceLengths)
 {
   // Ensure that the network is configured correctly.
@@ -512,7 +512,7 @@ typename MatType::elem_type RNN<
   // This will store the outputs of the network at each time step.  Note that we
   // only need to store `effectiveBPTTSteps` of output.  We will treat `outputs`
   // as a circular buffer.
-  arma::Cube<typename MatType::elem_type> outputs(
+  CubeType outputs(
       network.network.OutputSize(), batchSize, effectiveBPTTSteps);
 
   MatType stepData, outputData, responseData;
@@ -654,8 +654,8 @@ void RNN<
     InitializationRuleType,
     MatType
 >::ResetData(
-    arma::Cube<typename MatType::elem_type> predictors,
-    arma::Cube<typename MatType::elem_type> responses,
+    CubeType predictors,
+    CubeType responses,
     arma::urowvec sequenceLengths)
 {
   this->predictors = std::move(predictors);

mlpack/include/mlpack/methods/bayesian_linear_regression/bayesian_linear_regression_impl.hpp

@@ -273,11 +273,8 @@ inline double BayesianLinearRegression<ModelMatType>::CenterScaleData(
 {
   if (!centerData && !scaleData)
   {
-    dataProc = MatType(const_cast<ElemType*>(data.memptr()), data.n_rows,
-                                             data.n_cols, false, true);
-    responsesProc = ResponsesType(const_cast<ElemType*>(responses.memptr()),
-                                                        responses.n_elem, false,
-                                                        true);
+    MakeAlias(dataProc, data, data.n_rows, data.n_cols, 0, true);
+    MakeAlias(responsesProc, responses, responses.n_elem, 0, true);
   }
   else if (centerData && !scaleData)
   {
@@ -290,9 +287,7 @@ inline double BayesianLinearRegression<ModelMatType>::CenterScaleData(
   {
     dataScale = stddev(data, 0, 1);
     dataProc = data.each_col() / dataScale;
-    responsesProc = ResponsesType(const_cast<ElemType*>(responses.memptr()),
-                                                        responses.n_elem, false,
-                                                        true);
+    MakeAlias(responsesProc, responses, responses.n_elem, 0, true);
   }
   else
   {

mlpack/include/mlpack/methods/decision_tree/fitness_functions/gini_gain.hpp

@@ -68,14 +68,11 @@ class GiniGain
 
     // Count the number of elements in each class.  Use four auxiliary vectors
     // to exploit SIMD instructions if possible.
-    arma::vec countSpace(4 * numClasses);
-    arma::vec counts(countSpace.memptr(), numClasses, false, true);
-    arma::vec counts2(countSpace.memptr() + numClasses, numClasses, false,
-        true);
-    arma::vec counts3(countSpace.memptr() + 2 * numClasses, numClasses, false,
-        true);
-    arma::vec counts4(countSpace.memptr() + 3 * numClasses, numClasses, false,
-        true);
+    arma::vec countSpace(4 * numClasses), counts, counts2, counts3, counts4;
+    MakeAlias(counts, countSpace, numClasses, 0);
+    MakeAlias(counts2, countSpace, numClasses, numClasses);
+    MakeAlias(counts3, countSpace, numClasses, 2 * numClasses);
+    MakeAlias(counts4, countSpace, numClasses, 3 * numClasses);
 
     // Calculate the Gini impurity of the un-split node.
     double impurity = 0.0;

mlpack/include/mlpack/methods/decision_tree/fitness_functions/information_gain.hpp

@@ -69,14 +69,11 @@ class InformationGain
 
     // Count the number of elements in each class.  Use four auxiliary vectors
     // to exploit SIMD instructions if possible.
-    arma::vec countSpace(4 * numClasses);
-    arma::vec counts(countSpace.memptr(), numClasses, false, true);
-    arma::vec counts2(countSpace.memptr() + numClasses, numClasses, false,
-        true);
-    arma::vec counts3(countSpace.memptr() + 2 * numClasses, numClasses, false,
-        true);
-    arma::vec counts4(countSpace.memptr() + 3 * numClasses, numClasses, false,
-        true);
+    arma::vec countSpace(4 * numClasses), counts, counts2, counts3, counts4;
+    MakeAlias(counts, countSpace, numClasses, 0);
+    MakeAlias(counts2, countSpace, numClasses, numClasses);
+    MakeAlias(counts3, countSpace, numClasses, 2 * numClasses);
+    MakeAlias(counts4, countSpace, numClasses, 3 * numClasses);
 
     if (UseWeights)
     {

mlpack/include/mlpack/methods/gmm/diagonal_gmm_impl.hpp

@@ -172,7 +172,8 @@ inline void DiagonalGMM::LogProbability(const arma::mat& observation,
   // Assign value to the matrix.
   for (size_t i = 0; i < gaussians; i++)
   {
-    arma::vec temp(logProb.colptr(i), observation.n_cols, false, true);
+    arma::vec temp;
+    MakeAlias(temp, logProb, observation.n_cols, logProb.n_rows * i);
     dists[i].LogProbability(observation, temp);
   }
 

mlpack/include/mlpack/methods/gmm/eigenvalue_ratio_constraint.hpp

@@ -32,10 +32,10 @@ class EigenvalueRatioConstraint
    * which is the largest eigenvalue, so the first element of the vector should
    * be 1.  In addition, all other elements should be less than or equal to 1.
    */
-  EigenvalueRatioConstraint(const arma::vec& ratios) :
-      // Make an alias of the ratios vector.  It will never be modified here.
-      ratios(const_cast<double*>(ratios.memptr()), ratios.n_elem, false)
+  EigenvalueRatioConstraint(const arma::vec& ratios)
   {
+    // Make an alias of the ratios vector.  It will never be modified here.
+    MakeAlias(const_cast<arma::vec&>(this->ratios), ratios, ratios.n_elem);
     // Check validity of ratios.
     if (std::abs(ratios[0] - 1.0) > 1e-20)
       Log::Fatal << "EigenvalueRatioConstraint::EigenvalueRatioConstraint(): "

mlpack/include/mlpack/methods/gmm/gmm_impl.hpp

@@ -89,7 +89,8 @@ inline void GMM::LogProbability(const arma::mat& observation,
   // Assign value to the matrix.
   for (size_t i = 0; i < gaussians; i++)
   {
-    arma::vec temp(logProb.colptr(i), observation.n_cols, false, true);
+    arma::vec temp;
+    MakeAlias(temp, logProb, observation.n_cols, logProb.n_rows * i);
     dists[i].LogProbability(observation, temp);
   }
 

mlpack/include/mlpack/methods/hmm/hmm_impl.hpp

@@ -157,7 +157,8 @@ double HMM<Distribution>::Train(const std::vector<arma::mat>& dataSeq)
       for (size_t i = 0; i < logTransition.n_rows; i++)
       {
         // Define alias of desired column.
-        arma::vec alias(logProbs.colptr(i), logProbs.n_rows, false, true);
+        arma::vec alias;
+        MakeAlias(alias, logProbs, logProbs.n_rows, logProbs.n_rows * i);
         // Use advanced constructor for using logProbs directly.
         emission[i].LogProbability(dataSeq[seq], alias);
       }
@@ -358,7 +359,8 @@ double HMM<Distribution>::LogEstimate(const arma::mat& dataSeq,
   for (size_t i = 0; i < logTransition.n_rows; i++)
   {
     // Define alias of desired column.
-    arma::vec alias(logProbs.colptr(i), logProbs.n_rows, false, true);
+    arma::vec alias;
+    MakeAlias(alias, logProbs, logProbs.n_rows, logProbs.n_rows * i);
     // Use advanced constructor for using logProbs directly.
     emission[i].LogProbability(dataSeq, alias);
   }
@@ -515,7 +517,8 @@ double HMM<Distribution>::Predict(const arma::mat& dataSeq,
   for (size_t i = 0; i < logTransition.n_rows; i++)
   {
     // Define alias of desired column.
-    arma::vec alias(logProbs.colptr(i), logProbs.n_rows, false, true);
+    arma::vec alias;
+    MakeAlias(alias, logProbs, logProbs.n_rows, logProbs.n_rows * i);
     // Use advanced constructor for using logProbs directly.
     emission[i].LogProbability(dataSeq, alias);
   }
@@ -562,7 +565,8 @@ double HMM<Distribution>::LogLikelihood(const arma::mat& dataSeq) const
   for (size_t i = 0; i < logTransition.n_rows; i++)
   {
     // Define alias of desired column.
-    arma::vec alias(logProbs.colptr(i), logProbs.n_rows, false, true);
+    arma::vec alias;
+    MakeAlias(alias, logProbs, logProbs.n_rows, logProbs.n_rows * i);
     // Use advanced constructor for using logProbs directly.
     emission[i].LogProbability(dataSeq, alias);
   }
@@ -664,7 +668,8 @@ void HMM<Distribution>::Filter(const arma::mat& dataSeq,
   for (size_t i = 0; i < logTransition.n_rows; i++)
   {
     // Define alias of desired column.
-    arma::vec alias(logProbs.colptr(i), logProbs.n_rows, false, true);
+    arma::vec alias;
+    MakeAlias(alias, logProbs, logProbs.n_rows, logProbs.n_rows * i);
     // Use advanced constructor for using logProbs directly.
     emission[i].LogProbability(dataSeq, alias);
   }

mlpack/include/mlpack/methods/random_forest/random_forest.hpp

@@ -73,9 +73,9 @@ class RandomForest
    * @param dimensionSelector Instantiated dimension selection policy.
    * @param bootstrap Instantiated bootstrap policy.
    */
-  template<typename MatType>
+  template<typename MatType, typename LabelsType>
   RandomForest(const MatType& dataset,
-               const arma::Row<size_t>& labels,
+               const LabelsType& labels,
                const size_t numClasses,
                const size_t numTrees = 20,
                const size_t minimumLeafSize = 1,
@@ -104,10 +104,10 @@ class RandomForest
    * @param dimensionSelector Instantiated dimension selection policy.
    * @param bootstrap Instantiated bootstrap policy.
    */
-  template<typename MatType>
+  template<typename MatType, typename LabelsType>
   RandomForest(const MatType& dataset,
                const data::DatasetInfo& datasetInfo,
-               const arma::Row<size_t>& labels,
+               const LabelsType& labels,
                const size_t numClasses,
                const size_t numTrees = 20,
                const size_t minimumLeafSize = 1,
@@ -133,18 +133,22 @@ class RandomForest
    * @param dimensionSelector Instantiated dimension selection policy.
    * @param bootstrap Instantiated bootstrap policy.
    */
-  template<typename MatType>
+  template<typename MatType,
+           typename LabelsType,
+           typename WeightsType>
   RandomForest(const MatType& dataset,
-               const arma::Row<size_t>& labels,
+               const LabelsType& labels,
                const size_t numClasses,
-               const arma::rowvec& weights,
+               const WeightsType& weights,
                const size_t numTrees = 20,
                const size_t minimumLeafSize = 1,
                const double minimumGainSplit = 1e-7,
                const size_t maximumDepth = 0,
                DimensionSelectionType dimensionSelector =
                    DimensionSelectionType(),
-               BootstrapType bootstrap = BootstrapType());
+               BootstrapType bootstrap = BootstrapType(),
+               const std::enable_if_t<arma::is_arma_type<
+                   std::remove_reference_t<WeightsType>>::value>* = 0);
 
   /**
    * Create a random forest, training on the given weighted labeled training
@@ -166,19 +170,23 @@ class RandomForest
    * @param dimensionSelector Instantiated dimension selection policy.
    * @param bootstrap Instantiated bootstrap policy.
    */
-  template<typename MatType>
+  template<typename MatType,
+           typename LabelsType,
+           typename WeightsType>
   RandomForest(const MatType& dataset,
                const data::DatasetInfo& datasetInfo,
-               const arma::Row<size_t>& labels,
+               const LabelsType& labels,
                const size_t numClasses,
-               const arma::rowvec& weights,
+               const WeightsType& weights,
                const size_t numTrees = 20,
                const size_t minimumLeafSize = 1,
                const double minimumGainSplit = 1e-7,
                const size_t maximumDepth = 0,
                DimensionSelectionType dimensionSelector =
                    DimensionSelectionType(),
-               BootstrapType bootstrap = BootstrapType());
+               BootstrapType bootstrap = BootstrapType(),
+               const std::enable_if_t<arma::is_arma_type<
+                   std::remove_reference_t<WeightsType>>::value>* = 0);
 
   /**
    * Train the random forest on the given labeled training data with the given
@@ -200,9 +208,9 @@ class RandomForest
    * @param bootstrap Instantiated bootstrap policy.
    * @return The average entropy of all the decision trees trained under forest.
    */
-  template<typename MatType>
+  template<typename MatType, typename LabelsType>
   double Train(const MatType& data,
-               const arma::Row<size_t>& labels,
+               const LabelsType& labels,
                const size_t numClasses,
                const size_t numTrees = 20,
                const size_t minimumLeafSize = 1,
@@ -236,10 +244,10 @@ class RandomForest
    * @param bootstrap Instantiated bootstrap policy.
    * @return The average entropy of all the decision trees trained under forest.
    */
-  template<typename MatType>
+  template<typename MatType, typename LabelsType>
   double Train(const MatType& data,
                const data::DatasetInfo& datasetInfo,
-               const arma::Row<size_t>& labels,
+               const LabelsType& labels,
                const size_t numClasses,
                const size_t numTrees = 20,
                const size_t minimumLeafSize = 1,
@@ -271,11 +279,13 @@ class RandomForest
    * @param bootstrap Instantiated bootstrap policy.
    * @return The average entropy of all the decision trees trained under forest.
    */
-  template<typename MatType>
+  template<typename MatType,
+           typename LabelsType,
+           typename WeightsType>
   double Train(const MatType& data,
-               const arma::Row<size_t>& labels,
+               const LabelsType& labels,
                const size_t numClasses,
-               const arma::rowvec& weights,
+               const WeightsType& weights,
                const size_t numTrees = 20,
                const size_t minimumLeafSize = 1,
                const double minimumGainSplit = 1e-7,
@@ -283,7 +293,9 @@ class RandomForest
                const bool warmStart = false,
                DimensionSelectionType dimensionSelector =
                    DimensionSelectionType(),
-               BootstrapType bootstrap = BootstrapType());
+               BootstrapType bootstrap = BootstrapType(),
+               const std::enable_if_t<arma::is_arma_type<
+                   std::remove_reference_t<WeightsType>>::value>* = 0);
 
   /**
    * Train the random forest on the given weighted labeled training data with
@@ -308,12 +320,14 @@ class RandomForest
    * @param bootstrap Instantiated bootstrap policy.
    * @return The average entropy of all the decision trees trained under forest.
    */
-  template<typename MatType>
+  template<typename MatType,
+           typename LabelsType,
+           typename WeightsType>
   double Train(const MatType& data,
                const data::DatasetInfo& datasetInfo,
-               const arma::Row<size_t>& labels,
+               const LabelsType& labels,
                const size_t numClasses,
-               const arma::rowvec& weights,
+               const WeightsType& weights,
                const size_t numTrees = 20,
                const size_t minimumLeafSize = 1,
                const double minimumGainSplit = 1e-7,
@@ -321,7 +335,9 @@ class RandomForest
                const bool warmStart = false,
                DimensionSelectionType dimensionSelector =
                    DimensionSelectionType(),
-               BootstrapType bootstrap = BootstrapType());
+               BootstrapType bootstrap = BootstrapType(),
+               const std::enable_if_t<arma::is_arma_type<
+                   std::remove_reference_t<WeightsType>>::value>* = 0);
 
   /**
    * Predict the class of the given point.  If the random forest has not been
@@ -414,19 +430,23 @@ class RandomForest
    * @tparam MatType The type of data matrix (i.e. arma::mat).
    * @return The average entropy of all the decision trees trained under forest.
    */
-  template<bool UseWeights, bool UseDatasetInfo, typename MatType>
-  double Train(const MatType& data,
-               const data::DatasetInfo& datasetInfo,
-               const arma::Row<size_t>& labels,
-               const size_t numClasses,
-               const arma::rowvec& weights,
-               const size_t numTrees,
-               const size_t minimumLeafSize,
-               const double minimumGainSplit,
-               const size_t maximumDepth,
-               const bool warmStart,
-               DimensionSelectionType& dimensionSelector,
-               BootstrapType& bootstrap);
+  template<bool UseWeights,
+           bool UseDatasetInfo,
+           typename MatType,
+           typename LabelsType,
+           typename WeightsType>
+  double TrainInternal(const MatType& data,
+                       const data::DatasetInfo& datasetInfo,
+                       const LabelsType& labels,
+                       const size_t numClasses,
+                       const WeightsType& weights,
+                       const size_t numTrees,
+                       const size_t minimumLeafSize,
+                       const double minimumGainSplit,
+                       const size_t maximumDepth,
+                       const bool warmStart,
+                       DimensionSelectionType& dimensionSelector,
+                       BootstrapType& bootstrap);
 
   //! The trees in the forest.
   std::vector<DecisionTreeType> trees;