Use 'Type' and typedef conventions for loss functions.

src/mlpack/methods/ann/ffn.hpp

@@ -57,7 +57,7 @@ template<
  *     layer.
  */
 template<
-    typename OutputLayerType = NegativeLogLikelihood<>,
+    typename OutputLayerType = NegativeLogLikelihood,
     typename InitializationRuleType = RandomInitialization,
     typename MatType = arma::mat>
 class FFN

src/mlpack/methods/ann/loss_functions/binary_cross_entropy_loss.hpp

@@ -27,7 +27,7 @@ namespace ann /** Artificial Neural Network. */ {
  *         arma::sp_mat or arma::cube).
  */
 template<typename MatType = arma::mat>
-class BCELoss
+class BCELossType
 {
  public:
   /**
@@ -38,7 +38,7 @@ class BCELoss
    * @param reduction Reduction type. If true, it returns the mean of 
    *                  the loss. Else, it returns the sum.
    */
-  BCELoss(const double eps = 1e-10, const bool reduction = true);
+  BCELossType(const double eps = 1e-10, const bool reduction = true);
 
   /**
    * Computes the cross-entropy function.
@@ -84,13 +84,17 @@ class BCELoss
 
   //! Reduction type. If true, performs mean of loss else sum.
   bool reduction;
-}; // class BCELoss
+}; // class BCELossType
+
+typedef BCELossType<arma::mat> BCELoss;
 
 /**
- * Adding alias of BCELoss.
+ * Alias of BCELossType.
  */
+typedef BCELossType<arma::mat> CrossEntropyError;
+
 template<typename MatType = arma::mat>
-using CrossEntropyError = BCELoss<MatType>;
+using CrossEntropyErrorType = BCELossType<MatType>;
 
 } // namespace ann
 } // namespace mlpack

src/mlpack/methods/ann/loss_functions/binary_cross_entropy_loss_impl.hpp

@@ -19,14 +19,14 @@ namespace mlpack {
 namespace ann /** Artificial Neural Network. */ {
 
 template<typename MatType>
-BCELoss<MatType>::BCELoss(
+BCELossType<MatType>::BCELossType(
     const double eps, const bool reduction) : eps(eps), reduction(reduction)
 {
   // Nothing to do here.
 }
 
 template<typename MatType>
-typename MatType::elem_type BCELoss<MatType>::Forward(
+typename MatType::elem_type BCELossType<MatType>::Forward(
     const MatType& prediction,
     const MatType& target)
 {
@@ -40,7 +40,7 @@ typename MatType::elem_type BCELoss<MatType>::Forward(
 }
 
 template<typename MatType>
-void BCELoss<MatType>::Backward(
+void BCELossType<MatType>::Backward(
     const MatType& prediction,
     const MatType& target,
     MatType& loss)
@@ -52,7 +52,7 @@ void BCELoss<MatType>::Backward(
 
 template<typename MatType>
 template<typename Archive>
-void BCELoss<MatType>::serialize(
+void BCELossType<MatType>::serialize(
     Archive& ar,
     const uint32_t /* version */)
 {

src/mlpack/methods/ann/loss_functions/cosine_embedding_loss.hpp

@@ -33,11 +33,11 @@ namespace ann /** Artificial Neural Network. */ {
  *         arma::sp_mat or arma::cube).
  */
 template<typename MatType = arma::mat>
-class CosineEmbeddingLoss
+class CosineEmbeddingLossType
 {
  public:
   /**
-   * Create the CosineEmbeddingLoss object.
+   * Create the CosineEmbeddingLossType object.
    *
    * @param margin Increases cosine distance in case of dissimilarity.
    *               Refer definition of cosine-embedding-loss above.
@@ -47,7 +47,7 @@ class CosineEmbeddingLoss
    *                 Specifies reduction method i.e. sum or mean corresponding
    *                 to 0 and 1 respectively. Default value = 0.
    */
-  CosineEmbeddingLoss(const double margin = 0.0,
+  CosineEmbeddingLossType(const double margin = 0.0,
                       const bool similarity = true,
                       const bool takeMean = false);
 
@@ -103,7 +103,9 @@ class CosineEmbeddingLoss
 
   //! Locally-stored value of takeMean hyper-parameter.
   bool takeMean;
-}; // class CosineEmbeddingLoss
+}; // class CosineEmbeddingLossType
+
+typedef CosineEmbeddingLossType<arma::mat> CosineEmbeddingLoss;
 
 } // namespace ann
 } // namespace mlpack

src/mlpack/methods/ann/loss_functions/cosine_embedding_loss_impl.hpp

@@ -19,7 +19,7 @@ namespace mlpack {
 namespace ann /** Artificial Neural Network. */ {
 
 template<typename MatType>
-CosineEmbeddingLoss<MatType>::CosineEmbeddingLoss(
+CosineEmbeddingLossType<MatType>::CosineEmbeddingLossType(
     const double margin, const bool similarity, const bool takeMean):
     margin(margin), similarity(similarity), takeMean(takeMean)
 {
@@ -27,7 +27,7 @@ CosineEmbeddingLoss<MatType>::CosineEmbeddingLoss(
 }
 
 template<typename MatType>
-typename MatType::elem_type CosineEmbeddingLoss<MatType>::Forward(
+typename MatType::elem_type CosineEmbeddingLossType<MatType>::Forward(
     const MatType& prediction,
     const MatType& target)
 {
@@ -63,7 +63,7 @@ typename MatType::elem_type CosineEmbeddingLoss<MatType>::Forward(
 }
 
 template<typename MatType>
-void CosineEmbeddingLoss<MatType>::Backward(
+void CosineEmbeddingLossType<MatType>::Backward(
     const MatType& prediction,
     const MatType& target,
     MatType& loss)
@@ -101,7 +101,7 @@ void CosineEmbeddingLoss<MatType>::Backward(
 
 template<typename MatType>
 template<typename Archive>
-void CosineEmbeddingLoss<MatType>::serialize(
+void CosineEmbeddingLossType<MatType>::serialize(
     Archive& ar, const uint32_t /* version */)
 {
   ar(CEREAL_NVP(margin));

src/mlpack/methods/ann/loss_functions/dice_loss.hpp

@@ -44,15 +44,15 @@ namespace ann /** Artificial Neural Network. */ {
  *         arma::sp_mat or arma::cube).
  */
 template<typename MatType = arma::mat>
-class DiceLoss
+class DiceLossType
 {
  public:
   /**
-   * Create the DiceLoss object.
+   * Create the DiceLossType object.
    *
    * @param smooth The Laplace smoothing parameter.
    */
-  DiceLoss(const double smooth = 1);
+  DiceLossType(const double smooth = 1);
 
   /**
    * Computes the dice loss function.
@@ -90,7 +90,9 @@ class DiceLoss
  private:
   //! The parameter to avoid overfitting.
   double smooth;
-}; // class DiceLoss
+}; // class DiceLossType
+
+typedef DiceLossType<arma::mat> DiceLoss;
 
 } // namespace ann
 } // namespace mlpack

src/mlpack/methods/ann/loss_functions/dice_loss_impl.hpp

@@ -19,13 +19,13 @@ namespace mlpack {
 namespace ann /** Artificial Neural Network. */ {
 
 template<typename MatType>
-DiceLoss<MatType>::DiceLoss(const double smooth) : smooth(smooth)
+DiceLossType<MatType>::DiceLossType(const double smooth) : smooth(smooth)
 {
   // Nothing to do here.
 }
 
 template<typename MatType>
-typename MatType::elem_type DiceLoss<MatType>::Forward(
+typename MatType::elem_type DiceLossType<MatType>::Forward(
     const MatType& prediction,
     const MatType& target)
 {
@@ -35,7 +35,7 @@ typename MatType::elem_type DiceLoss<MatType>::Forward(
 }
 
 template<typename MatType>
-void DiceLoss<MatType>::Backward(
+void DiceLossType<MatType>::Backward(
     const MatType& prediction,
     const MatType& target,
     MatType& loss)
@@ -49,7 +49,7 @@ void DiceLoss<MatType>::Backward(
 
 template<typename MatType>
 template<typename Archive>
-void DiceLoss<MatType>::serialize(
+void DiceLossType<MatType>::serialize(
     Archive& ar,
     const uint32_t /* version */)
 {

src/mlpack/methods/ann/loss_functions/earth_mover_distance.hpp

@@ -27,13 +27,13 @@ namespace ann /** Artificial Neural Network. */ {
  *         arma::sp_mat or arma::cube).
  */
 template<typename MatType = arma::mat>
-class EarthMoverDistance
+class EarthMoverDistanceType
 {
  public:
   /**
-   * Create the EarthMoverDistance object.
+   * Create the EarthMoverDistanceType object.
    */
-  EarthMoverDistance();
+  EarthMoverDistanceType();
 
   /**
    * Ordinary feed forward pass of a neural network.
@@ -61,8 +61,10 @@ class EarthMoverDistance
    * Serialize the layer.
    */
   template<typename Archive>
-  void serialize(Archive& ar, const uint32_t /* version */);
-}; // class EarthMoverDistance
+  void serialize(Archive& ar, const uint32_t /* version */) { }
+}; // class EarthMoverDistanceType
+
+typedef EarthMoverDistanceType<arma::mat> EarthMoverDistance;
 
 } // namespace ann
 } // namespace mlpack

src/mlpack/methods/ann/loss_functions/earth_mover_distance_impl.hpp

@@ -19,13 +19,13 @@ namespace mlpack {
 namespace ann /** Artificial Neural Network. */ {
 
 template<typename MatType>
-EarthMoverDistance<MatType>::EarthMoverDistance()
+EarthMoverDistanceType<MatType>::EarthMoverDistanceType()
 {
   // Nothing to do here.
 }
 
 template<typename MatType>
-typename MatType::elem_type EarthMoverDistance<MatType>::Forward(
+typename MatType::elem_type EarthMoverDistanceType<MatType>::Forward(
     const MatType& prediction,
     const MatType& target)
 {
@@ -33,7 +33,7 @@ typename MatType::elem_type EarthMoverDistance<MatType>::Forward(
 }
 
 template<typename MatType>
-void EarthMoverDistance<MatType>::Backward(
+void EarthMoverDistanceType<MatType>::Backward(
     const MatType& /* prediction */,
     const MatType& target,
     MatType& loss)
@@ -41,15 +41,6 @@ void EarthMoverDistance<MatType>::Backward(
   loss = -target;
 }
 
-template<typename MatType>
-template<typename Archive>
-void EarthMoverDistance<MatType>::serialize(
-    Archive& /* ar */,
-    const uint32_t /* version */)
-{
-  /* Nothing to do here */
-}
-
 } // namespace ann
 } // namespace mlpack
 

src/mlpack/methods/ann/loss_functions/empty_loss.hpp

@@ -29,13 +29,13 @@ namespace ann /** Artificial Neural Network. */ {
  *         arma::sp_mat or arma::cube).
  */
 template<typename MatType = arma::mat>
-class EmptyLoss
+class EmptyLossType
 {
  public:
   /**
-   * Create the EmptyLoss object.
+   * Create the EmptyLossType object.
    */
-  EmptyLoss();
+  EmptyLossType();
 
   /**
    * Computes the Empty loss function.
@@ -58,10 +58,12 @@ class EmptyLoss
                 const MatType& target,
                 MatType& loss);
 
-  //! Serialize the EmptyLoss.
+  //! Serialize the EmptyLossType.
   template<typename Archive>
   void serialize(Archive& ar, const uint32_t /* version */) { }
-}; // class EmptyLoss
+}; // class EmptyLossType
+
+typedef EmptyLossType<arma::mat> EmptyLoss;
 
 } // namespace ann
 } // namespace mlpack

src/mlpack/methods/ann/loss_functions/empty_loss_impl.hpp

@@ -21,20 +21,20 @@ namespace mlpack {
 namespace ann /** Artificial Neural Network. */ {
 
 template<typename MatType>
-EmptyLoss<MatType>::EmptyLoss()
+EmptyLossType<MatType>::EmptyLossType()
 {
   // Nothing to do here.
 }
 
 template<typename MatType>
-double EmptyLoss<MatType>::Forward(
+double EmptyLossType<MatType>::Forward(
     const MatType& /* prediction */, const MatType& /* target */)
 {
   return 0;
 }
 
 template<typename MatType>
-void EmptyLoss<MatType>::Backward(
+void EmptyLossType<MatType>::Backward(
     const MatType& /* prediction */,
     const MatType& target,
     MatType& loss)

src/mlpack/methods/ann/loss_functions/hinge_embedding_loss.hpp

@@ -30,13 +30,13 @@ namespace ann /** Artificial Neural Network. */ {
  *         arma::sp_mat or arma::cube).
  */
 template<typename MatType = arma::mat>
-class HingeEmbeddingLoss
+class HingeEmbeddingLossType
 {
  public:
   /**
    * Create the Hinge Embedding object.
    */
-  HingeEmbeddingLoss();
+  HingeEmbeddingLossType();
 
   /**
    * Computes the Hinge Embedding loss function.
@@ -65,7 +65,9 @@ class HingeEmbeddingLoss
    */
   template<typename Archive>
   void serialize(Archive& ar, const uint32_t /* version */) { }
-}; // class HingeEmbeddingLoss
+}; // class HingeEmbeddingLossType
+
+typedef HingeEmbeddingLossType<arma::mat> HingeEmbeddingLoss;
 
 } // namespace ann
 } // namespace mlpack

src/mlpack/methods/ann/loss_functions/hinge_embedding_loss_impl.hpp

@@ -20,13 +20,13 @@ namespace mlpack {
 namespace ann /** Artificial Neural Network. */ {
 
 template<typename MatType>
-HingeEmbeddingLoss<MatType>::HingeEmbeddingLoss()
+HingeEmbeddingLossType<MatType>::HingeEmbeddingLossType()
 {
   // Nothing to do here.
 }
 
 template<typename MatType>
-typename MatType::elem_type HingeEmbeddingLoss<MatType>::Forward(
+typename MatType::elem_type HingeEmbeddingLossType<MatType>::Forward(
     const MatType& prediction,
     const MatType& target)
 {
@@ -35,7 +35,7 @@ typename MatType::elem_type HingeEmbeddingLoss<MatType>::Forward(
 }
 
 template<typename MatType>
-void HingeEmbeddingLoss<MatType>::Backward(
+void HingeEmbeddingLossType<MatType>::Backward(
     const MatType& prediction,
     const MatType& target,
     MatType& loss)

src/mlpack/methods/ann/loss_functions/hinge_loss.hpp

@@ -31,11 +31,11 @@ namespace ann /** Artificial Neural Network. */ {
  *         arma::sp_mat or arma::cube).
  */
 template<typename MatType = arma::mat>
-class HingeLoss
+class HingeLossType
 {
  public:
   /**
-   * Create HingeLoss object.
+   * Create HingeLossType object.
    *
    * @param reduction Specifies the reduction to apply to the output. If false,
    *                  'mean' reduction is used, where sum of the output will be
@@ -43,7 +43,7 @@ class HingeLoss
    *                  true, 'sum' reduction is used and the output will be
    *                  summed. It is set to true by default.
    */
-  HingeLoss(const bool reduction = true);
+  HingeLossType(const bool reduction = true);
 
   /**
    * Computes the Hinge loss function.
@@ -81,7 +81,9 @@ class HingeLoss
  private:
   //! The boolean value that tells if reduction is sum or mean.
   bool reduction;
-}; // class HingeLoss
+}; // class HingeLossType
+
+typedef HingeLossType<arma::mat> HingeLoss;
 
 } // namespace ann
 } // namespace mlpack

src/mlpack/methods/ann/loss_functions/hinge_loss_impl.hpp

@@ -20,14 +20,14 @@ namespace mlpack {
 namespace ann /** Artificial Neural Network. */ {
 
 template<typename MatType>
-HingeLoss<MatType>::HingeLoss(const bool reduction):
+HingeLossType<MatType>::HingeLossType(const bool reduction):
   reduction(reduction)
 {
   // Nothing to do here.
 }
 
 template<typename MatType>
-typename MatType::elem_type HingeLoss<MatType>::Forward(
+typename MatType::elem_type HingeLossType<MatType>::Forward(
     const MatType& prediction,
     const MatType& target)
 {
@@ -45,7 +45,7 @@ typename MatType::elem_type HingeLoss<MatType>::Forward(
 }
 
 template<typename MatType>
-void HingeLoss<MatType>::Backward(
+void HingeLossType<MatType>::Backward(
     const MatType& prediction,
     const MatType& target,
     MatType& loss)
@@ -59,7 +59,7 @@ void HingeLoss<MatType>::Backward(
 
 template<typename MatType>
 template<typename Archive>
-void HingeLoss<MatType>::serialize(
+void HingeLossType<MatType>::serialize(
     Archive& ar,
     const uint32_t /* version */)
 {

src/mlpack/methods/ann/loss_functions/huber_loss.hpp

@@ -30,17 +30,17 @@ namespace ann /** Artificial Neural Network. */ {
  *         arma::sp_mat or arma::cube).
  */
 template<typename MatType = arma::mat>
-class HuberLoss
+class HuberLossType
 {
  public:
   /**
-   * Create the HuberLoss object.
+   * Create the HuberLossType object.
    *
    * @param delta The threshold value upto which squared error is followed and
    *              after which absolute error is considered.
    * @param mean If true then mean loss is computed otherwise sum.
    */
-  HuberLoss(const double delta = 1.0, const bool mean = true);
+  HuberLossType(const double delta = 1.0, const bool mean = true);
 
   /**
    * Computes the Huber Loss function.
@@ -86,7 +86,9 @@ class HuberLoss
 
   //! Reduction type. If true, performs mean of loss else sum.
   bool mean;
-}; // class HuberLoss
+}; // class HuberLossType
+
+typedef HuberLossType<arma::mat> HuberLoss;
 
 } // namespace ann
 } // namespace mlpack

src/mlpack/methods/ann/loss_functions/huber_loss_impl.hpp

@@ -19,7 +19,7 @@ namespace mlpack {
 namespace ann /** Artificial Neural Network. */ {
 
 template<typename MatType>
-HuberLoss<MatType>::HuberLoss(
+HuberLossType<MatType>::HuberLossType(
     const double delta,
     const bool mean):
     delta(delta),
@@ -29,7 +29,7 @@ HuberLoss<MatType>::HuberLoss(
 }
 
 template<typename MatType>
-typename MatType::elem_type HuberLoss<MatType>::Forward(
+typename MatType::elem_type HuberLossType<MatType>::Forward(
     const MatType& prediction,
     const MatType& target)
 {
@@ -45,7 +45,7 @@ typename MatType::elem_type HuberLoss<MatType>::Forward(
 }
 
 template<typename MatType>
-void HuberLoss<MatType>::Backward(
+void HuberLossType<MatType>::Backward(
     const MatType& prediction,
     const MatType& target,
     MatType& loss)
@@ -66,7 +66,7 @@ void HuberLoss<MatType>::Backward(
 
 template<typename MatType>
 template<typename Archive>
-void HuberLoss<MatType>::serialize(
+void HuberLossType<MatType>::serialize(
     Archive& ar,
     const uint32_t /* version */)
 {

src/mlpack/methods/ann/loss_functions/kl_divergence.hpp

@@ -39,7 +39,7 @@ namespace ann /** Artificial Neural Network. */ {
  *         arma::sp_mat or arma::cube).
  */
 template<typename MatType = arma::mat>
-class KLDivergence
+class KLDivergenceType
 {
  public:
   /**
@@ -48,7 +48,7 @@ class KLDivergence
    *
    * @param takeMean Boolean variable to specify whether to take mean or not.
    */
-  KLDivergence(const bool takeMean = false);
+  KLDivergenceType(const bool takeMean = false);
 
   /**
    * Computes the Kullback–Leibler divergence error function.
@@ -86,7 +86,9 @@ class KLDivergence
  private:
   //! Boolean variable for taking mean or not.
   bool takeMean;
-}; // class KLDivergence
+}; // class KLDivergenceType
+
+typedef KLDivergenceType<arma::mat> KLDivergence;
 
 } // namespace ann
 } // namespace mlpack

src/mlpack/methods/ann/loss_functions/kl_divergence_impl.hpp

@@ -20,14 +20,14 @@ namespace mlpack {
 namespace ann /** Artificial Neural Network. */ {
 
 template<typename MatType>
-KLDivergence<MatType>::KLDivergence(const bool takeMean) :
+KLDivergenceType<MatType>::KLDivergenceType(const bool takeMean) :
     takeMean(takeMean)
 {
   // Nothing to do here.
 }
 
 template<typename MatType>
-typename MatType::elem_type KLDivergence<MatType>::Forward(
+typename MatType::elem_type KLDivergenceType<MatType>::Forward(
     const MatType& prediction,
     const MatType& target)
 {
@@ -43,7 +43,7 @@ typename MatType::elem_type KLDivergence<MatType>::Forward(
 }
 
 template<typename MatType>
-void KLDivergence<MatType>::Backward(
+void KLDivergenceType<MatType>::Backward(
     const MatType& prediction,
     const MatType& target,
     MatType& loss)
@@ -61,7 +61,7 @@ void KLDivergence<MatType>::Backward(
 
 template<typename MatType>
 template<typename Archive>
-void KLDivergence<MatType>::serialize(
+void KLDivergenceType<MatType>::serialize(
     Archive& ar,
     const uint32_t /* version */)
 {

src/mlpack/methods/ann/loss_functions/l1_loss.hpp

@@ -27,16 +27,16 @@ namespace ann /** Artificial Neural Network. */ {
  *         arma::sp_mat or arma::cube).
  */
 template<typename MatType = arma::mat>
-class L1Loss
+class L1LossType
 {
  public:
   /**
-   * Create the L1Loss object.
+   * Create the L1LossType object.
    *
    * @param mean Reduction type. If true, it returns the mean of 
    * the loss. Else, it returns the sum.
    */
-  L1Loss(const bool mean = true);
+  L1LossType(const bool mean = true);
 
   /**
    * Computes the L1 Loss function.
@@ -74,7 +74,9 @@ class L1Loss
  private:
   //! Reduction type. If true, performs mean of loss else sum.
   bool mean;
-}; // class L1Loss
+}; // class L1LossType
+
+typedef L1LossType<arma::mat> L1Loss;
 
 } // namespace ann
 } // namespace mlpack

src/mlpack/methods/ann/loss_functions/l1_loss_impl.hpp

@@ -19,14 +19,14 @@ namespace mlpack {
 namespace ann /** Artificial Neural Network. */ {
 
 template<typename MatType>
-L1Loss<MatType>::L1Loss(const bool mean):
+L1LossType<MatType>::L1LossType(const bool mean):
   mean(mean)
 {
   // Nothing to do here.
 }
 
 template<typename MatType>
-typename MatType::elem_type L1Loss<MatType>::Forward(
+typename MatType::elem_type L1LossType<MatType>::Forward(
     const MatType& prediction,
     const MatType& target)
 {
@@ -37,7 +37,7 @@ typename MatType::elem_type L1Loss<MatType>::Forward(
 }
 
 template<typename MatType>
-void L1Loss<MatType>::Backward(
+void L1LossType<MatType>::Backward(
     const MatType& prediction,
     const MatType& target,
     MatType& loss)
@@ -47,7 +47,7 @@ void L1Loss<MatType>::Backward(
 
 template<typename MatType>
 template<typename Archive>
-void L1Loss<MatType>::serialize(
+void L1LossType<MatType>::serialize(
     Archive& ar,
     const uint32_t /* version */)
 {

src/mlpack/methods/ann/loss_functions/log_cosh_loss.hpp

@@ -29,7 +29,7 @@ namespace ann /** Artificial Neural Network. */ {
  *         arma::sp_mat or arma::cube).
  */
 template<typename MatType = arma::mat>
-class LogCoshLoss
+class LogCoshLossType
 {
  public:
   /**
@@ -43,7 +43,7 @@ class LogCoshLoss
    *          function more sensitive to small losses around the
    *          origin. Default value = 1.0.
    */
-  LogCoshLoss(const double a = 1.0);
+  LogCoshLossType(const double a = 1.0);
 
   /**
    * Computes the Log-Hyperbolic-Cosine loss function.
@@ -81,7 +81,9 @@ class LogCoshLoss
  private:
   //! Hyperparameter a for smoothening function curve.
   double a;
-}; // class LogCoshLoss
+}; // class LogCoshLossType
+
+typedef LogCoshLossType<arma::mat> LogCoshLoss;
 
 } // namespace ann
 } // namespace mlpack

src/mlpack/methods/ann/loss_functions/log_cosh_loss_impl.hpp

@@ -20,14 +20,14 @@ namespace mlpack {
 namespace ann /** Artificial Neural Network. */ {
 
 template<typename MatType>
-LogCoshLoss<MatType>::LogCoshLoss(const double a) :
+LogCoshLossType<MatType>::LogCoshLossType(const double a) :
     a(a)
 {
   Log::Assert(a > 0, "Hyper-Parameter \'a\' must be positive");
 }
 
 template<typename MatType>
-typename MatType::elem_type LogCoshLoss<MatType>::Forward(
+typename MatType::elem_type LogCoshLossType<MatType>::Forward(
     const MatType& prediction,
     const MatType& target)
 {
@@ -35,7 +35,7 @@ typename MatType::elem_type LogCoshLoss<MatType>::Forward(
 }
 
 template<typename MatType>
-void LogCoshLoss<MatType>::Backward(
+void LogCoshLossType<MatType>::Backward(
     const MatType& prediction,
     const MatType& target,
     MatType& loss)
@@ -45,7 +45,7 @@ void LogCoshLoss<MatType>::Backward(
 
 template<typename MatType>
 template<typename Archive>
-void LogCoshLoss<MatType>::serialize(
+void LogCoshLossType<MatType>::serialize(
     Archive& ar,
     const uint32_t /* version */)
 {

src/mlpack/methods/ann/loss_functions/margin_ranking_loss.hpp

@@ -29,15 +29,15 @@ namespace ann /** Artificial Neural Network. */ {
  *         arma::sp_mat or arma::cube).
  */
 template<typename MatType = arma::mat>
-class MarginRankingLoss
+class MarginRankingLossType
 {
  public:
   /**
-   * Create the MarginRankingLoss object with Hyperparameter margin.
+   * Create the MarginRankingLossType object with Hyperparameter margin.
    * Hyperparameter margin defines a minimum distance between correctly ranked
    * samples.
    */
-  MarginRankingLoss(const double margin = 1.0);
+  MarginRankingLossType(const double margin = 1.0);
 
   /**
    * Computes the Margin Ranking Loss function.
@@ -75,7 +75,9 @@ class MarginRankingLoss
  private:
   //! The margin value used in calculating Margin Ranking Loss.
   double margin;
-}; // class MarginRankingLoss
+}; // class MarginRankingLossType
+
+typedef MarginRankingLossType<arma::mat> MarginRankingLoss;
 
 } // namespace ann
 } // namespace mlpack

src/mlpack/methods/ann/loss_functions/margin_ranking_loss_impl.hpp

@@ -19,14 +19,14 @@ namespace mlpack {
 namespace ann /** Artifical Neural Network. */ {
 
 template<typename MatType>
-MarginRankingLoss<MatType>::MarginRankingLoss(
+MarginRankingLossType<MatType>::MarginRankingLossType(
     const double margin) : margin(margin)
 {
   // Nothing to do here.
 }
 
 template<typename MatType>
-typename MatType::elem_type MarginRankingLoss<MatType>::Forward(
+typename MatType::elem_type MarginRankingLossType<MatType>::Forward(
     const MatType& prediction,
     const MatType& target)
 {
@@ -40,7 +40,7 @@ typename MatType::elem_type MarginRankingLoss<MatType>::Forward(
 }
 
 template<typename MatType>
-void MarginRankingLoss<MatType>::Backward(
+void MarginRankingLossType<MatType>::Backward(
     const MatType& prediction,
     const MatType& target,
     MatType& loss)
@@ -58,7 +58,7 @@ void MarginRankingLoss<MatType>::Backward(
 
 template<typename MatType>
 template<typename Archive>
-void MarginRankingLoss<MatType>::serialize(
+void MarginRankingLossType<MatType>::serialize(
     Archive& ar,
     const uint32_t /* version */)
 {

src/mlpack/methods/ann/loss_functions/mean_absolute_percentage_error.hpp

@@ -43,13 +43,13 @@ namespace ann /** Artificial Neural Network. */ {
  *         arma::sp_mat or arma::cube).
  */
 template<typename MatType = arma::mat>
-class MeanAbsolutePercentageError
+class MeanAbsolutePercentageErrorType
 {
  public:
   /**
-   * Create the MeanAbsolutePercentageError object.
+   * Create the MeanAbsolutePercentageErrorType object.
    */
-  MeanAbsolutePercentageError();
+  MeanAbsolutePercentageErrorType();
 
   /**
    * Computes the mean absolute percentage error function.
@@ -78,7 +78,9 @@ class MeanAbsolutePercentageError
    */
   template<typename Archive>
   void serialize(Archive& ar, const unsigned int /* version */) { }
-}; // class MeanAbsolutePercentageError
+}; // class MeanAbsolutePercentageErrorType
+
+typedef MeanAbsolutePercentageErrorType<arma::mat> MeanAbsolutePercentageError;
 
 } // namespace ann
 } // namespace mlpack

src/mlpack/methods/ann/loss_functions/mean_absolute_percentage_error_impl.hpp

@@ -19,13 +19,13 @@ namespace mlpack {
 namespace ann /** Artificial Neural Network. */ {
 
 template<typename MatType>
-MeanAbsolutePercentageError<MatType>::MeanAbsolutePercentageError()
+MeanAbsolutePercentageErrorType<MatType>::MeanAbsolutePercentageErrorType()
 {
   // Nothing to do here.
 }
 
 template<typename MatType>
-typename MatType::elem_type MeanAbsolutePercentageError<MatType>::Forward(
+typename MatType::elem_type MeanAbsolutePercentageErrorType<MatType>::Forward(
     const MatType& prediction,
     const MatType& target)
 {
@@ -34,7 +34,7 @@ typename MatType::elem_type MeanAbsolutePercentageError<MatType>::Forward(
 }
 
 template<typename MatType>
-void MeanAbsolutePercentageError<MatType>::Backward(
+void MeanAbsolutePercentageErrorType<MatType>::Backward(
     const MatType& prediction,
     const MatType& target,
     MatType& loss)

src/mlpack/methods/ann/loss_functions/mean_bias_error.hpp

@@ -27,13 +27,13 @@ namespace ann /** Artificial Neural Network. */ {
  *         arma::sp_mat or arma::cube).
  */
 template<typename MatType = arma::mat>
-class MeanBiasError
+class MeanBiasErrorType
 {
  public:
   /**
-   * Create the MeanBiasError object.
+   * Create the MeanBiasErrorType object.
    */
-  MeanBiasError();
+  MeanBiasErrorType();
 
   /**
    * Computes the mean bias error function.
@@ -62,7 +62,9 @@ class MeanBiasError
    */
   template<typename Archive>
   void serialize(Archive& ar, const uint32_t /* version */);
-}; // class MeanBiasError
+}; // class MeanBiasErrorType
+
+typedef MeanBiasErrorType<arma::mat> MeanBiasError;
 
 } // namespace ann
 } // namespace mlpack

src/mlpack/methods/ann/loss_functions/mean_bias_error_impl.hpp

@@ -20,13 +20,13 @@ namespace mlpack {
 namespace ann /** Artificial Neural Network. */ {
 
 template<typename MatType>
-MeanBiasError<MatType>::MeanBiasError()
+MeanBiasErrorType<MatType>::MeanBiasErrorType()
 {
   // Nothing to do here.
 }
 
 template<typename MatType>
-typename MatType::elem_type MeanBiasError<MatType>::Forward(
+typename MatType::elem_type MeanBiasErrorType<MatType>::Forward(
     const MatType& prediction,
     const MatType& target)
 {
@@ -34,7 +34,7 @@ typename MatType::elem_type MeanBiasError<MatType>::Forward(
 }
 
 template<typename MatType>
-void MeanBiasError<MatType>::Backward(
+void MeanBiasErrorType<MatType>::Backward(
     const MatType& prediction,
     const MatType& /* target */,
     MatType& loss)

src/mlpack/methods/ann/loss_functions/mean_squared_error.hpp

@@ -28,13 +28,13 @@ namespace ann /** Artificial Neural Network. */ {
  *         arma::sp_mat or arma::cube).
  */
 template<typename MatType = arma::mat>
-class MeanSquaredError
+class MeanSquaredErrorType
 {
  public:
   /**
-   * Create the MeanSquaredError object.
+   * Create the MeanSquaredErrorType object.
    */
-  MeanSquaredError();
+  MeanSquaredErrorType();
 
   /**
    * Computes the mean squared error function.
@@ -63,7 +63,9 @@ class MeanSquaredError
    */
   template<typename Archive>
   void serialize(Archive& ar, const uint32_t /* version */) { }
-}; // class MeanSquaredError
+}; // class MeanSquaredErrorType
+
+typedef MeanSquaredErrorType<arma::mat> MeanSquaredError;
 
 } // namespace ann
 } // namespace mlpack

src/mlpack/methods/ann/loss_functions/mean_squared_error_impl.hpp

@@ -19,13 +19,13 @@ namespace mlpack {
 namespace ann /** Artificial Neural Network. */ {
 
 template<typename MatType>
-MeanSquaredError<MatType>::MeanSquaredError()
+MeanSquaredErrorType<MatType>::MeanSquaredErrorType()
 {
   // Nothing to do here.
 }
 
 template<typename MatType>
-typename MatType::elem_type MeanSquaredError<MatType>::Forward(
+typename MatType::elem_type MeanSquaredErrorType<MatType>::Forward(
     const MatType& prediction,
     const MatType& target)
 {
@@ -33,7 +33,7 @@ typename MatType::elem_type MeanSquaredError<MatType>::Forward(
 }
 
 template<typename MatType>
-void MeanSquaredError<MatType>::Backward(
+void MeanSquaredErrorType<MatType>::Backward(
     const MatType& prediction,
     const MatType& target,
     MatType& loss)

src/mlpack/methods/ann/loss_functions/mean_squared_logarithmic_error.hpp

@@ -27,13 +27,13 @@ namespace ann /** Artificial Neural Network. */ {
  *         arma::sp_mat or arma::cube).
  */
 template<typename MatType = arma::mat>
-class MeanSquaredLogarithmicError
+class MeanSquaredLogarithmicErrorType
 {
  public:
   /**
-   * Create the MeanSquaredLogarithmicError object.
+   * Create the MeanSquaredLogarithmicErrorType object.
    */
-  MeanSquaredLogarithmicError();
+  MeanSquaredLogarithmicErrorType();
 
   /**
    * Computes the mean squared logarithmic error function.
@@ -62,7 +62,9 @@ class MeanSquaredLogarithmicError
    */
   template<typename Archive>
   void serialize(Archive& ar, const uint32_t /* version */) { }
-}; // class MeanSquaredLogarithmicError
+}; // class MeanSquaredLogarithmicErrorType
+
+typedef MeanSquaredLogarithmicErrorType<arma::mat> MeanSquaredLogarithmicError;
 
 } // namespace ann
 } // namespace mlpack

src/mlpack/methods/ann/loss_functions/mean_squared_logarithmic_error_impl.hpp

@@ -19,14 +19,13 @@ namespace mlpack {
 namespace ann /** Artificial Neural Network. */ {
 
 template<typename MatType>
-MeanSquaredLogarithmicError<MatType>
-::MeanSquaredLogarithmicError()
+MeanSquaredLogarithmicErrorType<MatType>::MeanSquaredLogarithmicErrorType()
 {
   // Nothing to do here.
 }
 
 template<typename MatType>
-typename MatType::elem_type MeanSquaredLogarithmicError<MatType>::Forward(
+typename MatType::elem_type MeanSquaredLogarithmicErrorType<MatType>::Forward(
     const MatType& prediction,
     const MatType& target)
 {
@@ -35,7 +34,7 @@ typename MatType::elem_type MeanSquaredLogarithmicError<MatType>::Forward(
 }
 
 template<typename MatType>
-void MeanSquaredLogarithmicError<MatType>::Backward(
+void MeanSquaredLogarithmicErrorType<MatType>::Backward(
     const MatType& prediction,
     const MatType& target,
     MatType& loss)

src/mlpack/methods/ann/loss_functions/multilabel_softmargin_loss.hpp

@@ -28,11 +28,11 @@ namespace ann /** Artificial Neural Network. */ {
  *         arma::sp_mat or arma::cube).
  */
 template<typename MatType = arma::mat>
-class MultiLabelSoftMarginLoss
+class MultiLabelSoftMarginLossType
 {
  public:
   /**
-   * Create the MultiLabelSoftMarginLoss object.
+   * Create the MultiLabelSoftMarginLossType object.
    *
    * @param reduction Specifies the reduction to apply to the output. If false,
    *                  'mean' reduction is used, where sum of the output will be
@@ -42,7 +42,7 @@ class MultiLabelSoftMarginLoss
    * @param weights A manual rescaling weight given to each class. It is a
    *                (1, numClasses) row vector.
    */
-  MultiLabelSoftMarginLoss(
+  MultiLabelSoftMarginLossType(
       const bool reduction = true,
       const arma::Row<typename MatType::elem_type>& weights =
           arma::Row<typename MatType::elem_type>());
@@ -98,7 +98,9 @@ class MultiLabelSoftMarginLoss
 
   // An internal parameter used during initialisation of class weights.
   bool weighted;
-}; // class MultiLabelSoftMarginLoss
+}; // class MultiLabelSoftMarginLossType
+
+typedef MultiLabelSoftMarginLossType<arma::mat> MultiLabelSoftMarginLoss;
 
 } // namespace ann
 } // namespace mlpack

src/mlpack/methods/ann/loss_functions/multilabel_softmargin_loss_impl.hpp

@@ -19,7 +19,7 @@ namespace mlpack {
 namespace ann /** Artifical Neural Network. */ {
 
 template<typename MatType>
-MultiLabelSoftMarginLoss<MatType>::MultiLabelSoftMarginLoss(
+MultiLabelSoftMarginLossType<MatType>::MultiLabelSoftMarginLossType(
     const bool reduction,
     const arma::Row<typename MatType::elem_type>& weights) :
     reduction(reduction),
@@ -33,7 +33,7 @@ MultiLabelSoftMarginLoss<MatType>::MultiLabelSoftMarginLoss(
 }
 
 template<typename MatType>
-typename MatType::elem_type MultiLabelSoftMarginLoss<MatType>::Forward(
+typename MatType::elem_type MultiLabelSoftMarginLossType<MatType>::Forward(
     const MatType& input, const MatType& target)
 {
   if (!weighted)
@@ -54,7 +54,7 @@ typename MatType::elem_type MultiLabelSoftMarginLoss<MatType>::Forward(
 }
 
 template<typename MatType>
-void MultiLabelSoftMarginLoss<MatType>::Backward(
+void MultiLabelSoftMarginLossType<MatType>::Backward(
     const MatType& input,
     const MatType& target,
     MatType& output)
@@ -70,7 +70,7 @@ void MultiLabelSoftMarginLoss<MatType>::Backward(
 
 template<typename MatType>
 template<typename Archive>
-void MultiLabelSoftMarginLoss<MatType>::serialize(
+void MultiLabelSoftMarginLossType<MatType>::serialize(
     Archive& ar,
     const unsigned int /* version */)
 {

src/mlpack/methods/ann/loss_functions/negative_log_likelihood.hpp

@@ -2,7 +2,7 @@
  * @file methods/ann/loss_functions/negative_log_likelihood.hpp
  * @author Marcus Edel
  *
- * Definition of the NegativeLogLikelihood class.
+ * Definition of the NegativeLogLikelihoodType class.
  *
  * mlpack is free software; you may redistribute it and/or modify it under the
  * terms of the 3-clause BSD license.  You should have received a copy of the
@@ -29,13 +29,13 @@ namespace ann /** Artificial Neural Network. */ {
  *         arma::sp_mat or arma::cube).
  */
 template<typename MatType = arma::mat>
-class NegativeLogLikelihood
+class NegativeLogLikelihoodType
 {
  public:
   /**
-   * Create the NegativeLogLikelihoodLayer object.
+   * Create the NegativeLogLikelihoodTypeLayer object.
    */
-  NegativeLogLikelihood();
+  NegativeLogLikelihoodType();
 
   /**
    * Computes the Negative log likelihood.
@@ -69,7 +69,9 @@ class NegativeLogLikelihood
    */
   template<typename Archive>
   void serialize(Archive& /* ar */, const uint32_t /* version */) { }
-}; // class NegativeLogLikelihood
+}; // class NegativeLogLikelihoodType
+
+typedef NegativeLogLikelihoodType<arma::mat> NegativeLogLikelihood;
 
 } // namespace ann
 } // namespace mlpack

src/mlpack/methods/ann/loss_functions/negative_log_likelihood_impl.hpp

@@ -2,7 +2,7 @@
  * @file methods/ann/loss_functions/negative_log_likelihood_impl.hpp
  * @author Marcus Edel
  *
- * Implementation of the NegativeLogLikelihood class.
+ * Implementation of the NegativeLogLikelihoodType class.
  *
  * mlpack is free software; you may redistribute it and/or modify it under the
  * terms of the 3-clause BSD license.  You should have received a copy of the
@@ -19,13 +19,13 @@ namespace mlpack {
 namespace ann /** Artificial Neural Network. */ {
 
 template<typename MatType>
-NegativeLogLikelihood<MatType>::NegativeLogLikelihood()
+NegativeLogLikelihoodType<MatType>::NegativeLogLikelihoodType()
 {
   // Nothing to do here.
 }
 
 template<typename MatType>
-double NegativeLogLikelihood<MatType>::Forward(
+double NegativeLogLikelihoodType<MatType>::Forward(
     const MatType& prediction,
     const MatType& target)
 {
@@ -42,7 +42,7 @@ double NegativeLogLikelihood<MatType>::Forward(
 }
 
 template<typename MatType>
-void NegativeLogLikelihood<MatType>::Backward(
+void NegativeLogLikelihoodType<MatType>::Backward(
       const MatType& prediction,
       const MatType& target,
       MatType& loss)

src/mlpack/methods/ann/loss_functions/poisson_nll_loss.hpp

@@ -2,7 +2,7 @@
  * @file methods/ann/loss_functions/poisson_nll_loss.hpp
  * @author Mrityunjay Tripathi
  *
- * Definition of the PoissonNLLLoss class. It is the negative log likelihood of
+ * Definition of the PoissonNLLLossType class. It is the negative log likelihood of
  * the Poisson distribution.
  *
  * mlpack is free software; you may redistribute it and/or modify it under the
@@ -30,11 +30,11 @@ namespace ann /** Artificial Neural Network. */ {
  *         arma::sp_mat or arma::cube).
  */
 template<typename MatType = arma::mat>
-class PoissonNLLLoss
+class PoissonNLLLossType
 {
  public:
   /**
-   * Create the PoissonNLLLoss object.
+   * Create the PoissonNLLLossType object.
    *
    * @param logInput If true the loss is computed as
    *        \f$ \exp(input) - target \cdot input \f$, if false then the loss is
@@ -44,7 +44,7 @@ class PoissonNLLLoss
    * @param eps A small value to prevent 0 in denominators and logarithms.
    * @param mean When true, mean loss is computed otherwise total loss.
    */
-  PoissonNLLLoss(const bool logInput = true,
+  PoissonNLLLossType(const bool logInput = true,
                  const bool full = false,
                  const typename MatType::elem_type eps = 1e-08,
                  const bool mean = true);
@@ -135,7 +135,9 @@ class PoissonNLLLoss
 
   //! Boolean value that tells if mean of the total loss has to be taken.
   bool mean;
-}; // class PoissonNLLLoss
+}; // class PoissonNLLLossType
+
+typedef PoissonNLLLossType<arma::mat> PoissonNLLLoss;
 
 } // namespace ann
 } // namespace mlpack

src/mlpack/methods/ann/loss_functions/poisson_nll_loss_impl.hpp

@@ -2,7 +2,7 @@
  * @file methods/ann/loss_functions/poisson_nll_loss_impl.hpp
  * @author Mrityunjay Tripathi
  *
- * Implementation of the PoissonNLLLoss class.
+ * Implementation of the PoissonNLLLossType class.
  *
  * mlpack is free software; you may redistribute it and/or modify it under the
  * terms of the 3-clause BSD license.  You should have received a copy of the
@@ -20,7 +20,7 @@ namespace mlpack {
 namespace ann /** Artificial Neural Network. */ {
 
 template<typename MatType>
-PoissonNLLLoss<MatType>::PoissonNLLLoss(
+PoissonNLLLossType<MatType>::PoissonNLLLossType(
     const bool logInput,
     const bool full,
     const typename MatType::elem_type eps,
@@ -34,7 +34,7 @@ PoissonNLLLoss<MatType>::PoissonNLLLoss(
 }
 
 template<typename MatType>
-typename MatType::elem_type PoissonNLLLoss<MatType>::Forward(
+typename MatType::elem_type PoissonNLLLossType<MatType>::Forward(
     const MatType& prediction,
     const MatType& target)
 {
@@ -60,7 +60,7 @@ typename MatType::elem_type PoissonNLLLoss<MatType>::Forward(
 }
 
 template<typename MatType>
-void PoissonNLLLoss<MatType>::Backward(
+void PoissonNLLLossType<MatType>::Backward(
     const MatType& prediction,
     const MatType& target,
     MatType& loss)
@@ -78,7 +78,7 @@ void PoissonNLLLoss<MatType>::Backward(
 
 template<typename MatType>
 template<typename Archive>
-void PoissonNLLLoss<MatType>::serialize(
+void PoissonNLLLossType<MatType>::serialize(
     Archive& ar,
     const uint32_t /* version */)
 {

src/mlpack/methods/ann/loss_functions/reconstruction_loss.hpp

@@ -33,13 +33,13 @@ template<
     typename MatType = arma::mat,
     typename DistType = BernoulliDistribution<MatType>
 >
-class ReconstructionLoss
+class ReconstructionLossType
 {
  public:
   /**
-   * Create the ReconstructionLoss object.
+   * Create the ReconstructionLossType object.
    */
-  ReconstructionLoss();
+  ReconstructionLossType();
 
   /**
    * Computes the reconstruction loss.
@@ -72,7 +72,9 @@ class ReconstructionLoss
  private:
   //! Locally-stored distribution object.
   DistType dist;
-}; // class ReconstructionLoss
+}; // class ReconstructionLossType
+
+typedef ReconstructionLossType<arma::mat> ReconstructionLoss;
 
 } // namespace ann
 } // namespace mlpack

src/mlpack/methods/ann/loss_functions/reconstruction_loss_impl.hpp

@@ -19,13 +19,13 @@ namespace mlpack {
 namespace ann /** Artificial Neural Network. */ {
 
 template<typename MatType, typename DistType>
-ReconstructionLoss<MatType, DistType>::ReconstructionLoss()
+ReconstructionLossType<MatType, DistType>::ReconstructionLossType()
 {
   // Nothing to do here.
 }
 
 template<typename MatType, typename DistType>
-typename MatType::elem_type ReconstructionLoss<MatType, DistType>::Forward(
+typename MatType::elem_type ReconstructionLossType<MatType, DistType>::Forward(
     const MatType& prediction, const MatType& target)
 {
   dist = DistType(prediction);
@@ -33,7 +33,7 @@ typename MatType::elem_type ReconstructionLoss<MatType, DistType>::Forward(
 }
 
 template<typename MatType, typename DistType>
-void ReconstructionLoss<MatType, DistType>::Backward(
+void ReconstructionLossType<MatType, DistType>::Backward(
     const MatType& /* prediction */,
     const MatType& target,
     MatType& loss)
@@ -44,7 +44,7 @@ void ReconstructionLoss<MatType, DistType>::Backward(
 
 template<typename MatType, typename DistType>
 template<typename Archive>
-void ReconstructionLoss<MatType, DistType>::serialize(
+void ReconstructionLossType<MatType, DistType>::serialize(
     Archive& ar,
     const uint32_t /* version */)
 {

src/mlpack/methods/ann/loss_functions/sigmoid_cross_entropy_error.hpp

@@ -19,7 +19,7 @@ namespace mlpack {
 namespace ann /** Artificial Neural Network. */ {
 
 /**
- * The SigmoidCrossEntropyError performance function measures the network's
+ * The SigmoidCrossEntropyErrorType performance function measures the network's
  * performance according to the cross-entropy function between the input and
  * target distributions. This function calculates the cross entropy
  * given the real values instead of providing the sigmoid activations.
@@ -46,13 +46,13 @@ namespace ann /** Artificial Neural Network. */ {
  *         arma::sp_mat or arma::cube).
  */
 template<typename MatType = arma::mat>
-class SigmoidCrossEntropyError
+class SigmoidCrossEntropyErrorType
 {
  public:
   /**
-   * Create the SigmoidCrossEntropyError object.
+   * Create the SigmoidCrossEntropyErrorType object.
    */
-  SigmoidCrossEntropyError();
+  SigmoidCrossEntropyErrorType();
 
   /**
    * Computes the Sigmoid CrossEntropy Error functions.
@@ -82,7 +82,9 @@ class SigmoidCrossEntropyError
    */
   template<typename Archive>
   void serialize(Archive& ar, const uint32_t /* version */) { }
-}; // class SigmoidCrossEntropy
+}; // class SigmoidCrossEntropyErrorType
+
+typedef SigmoidCrossEntropyErrorType<arma::mat> SigmoidCrossEntropyError;
 
 } // namespace ann
 } // namespace mlpack

src/mlpack/methods/ann/loss_functions/sigmoid_cross_entropy_error_impl.hpp

@@ -21,13 +21,14 @@ namespace mlpack {
 namespace ann /** Artificial Neural Network. */ {
 
 template<typename MatType>
-SigmoidCrossEntropyError<MatType>::SigmoidCrossEntropyError()
+SigmoidCrossEntropyErrorType<MatType>::SigmoidCrossEntropyErrorType()
 {
   // Nothing to do here.
 }
 
 template<typename MatType>
-inline typename MatType::elem_type SigmoidCrossEntropyError<MatType>::Forward(
+inline typename MatType::elem_type
+SigmoidCrossEntropyErrorType<MatType>::Forward(
     const MatType& prediction,
     const MatType& target)
 {
@@ -43,7 +44,7 @@ inline typename MatType::elem_type SigmoidCrossEntropyError<MatType>::Forward(
 }
 
 template<typename MatType>
-inline void SigmoidCrossEntropyError<MatType>::Backward(
+inline void SigmoidCrossEntropyErrorType<MatType>::Backward(
     const MatType& prediction,
     const MatType& target,
     MatType& loss)

src/mlpack/methods/ann/loss_functions/soft_margin_loss.hpp

@@ -28,11 +28,11 @@ namespace ann /** Artificial Neural Network. */ {
  *         arma::sp_mat or arma::cube).
  */
 template<typename MatType = arma::mat>
-class SoftMarginLoss
+class SoftMarginLossType
 {
  public:
   /**
-   * Create the SoftMarginLoss object.
+   * Create the SoftMarginLossType object.
    *
    * @param reduction Specifies the reduction to apply to the output. If false,
    *                  'mean' reduction is used, where sum of the output will be
@@ -40,7 +40,7 @@ class SoftMarginLoss
    *                  true, 'sum' reduction is used and the output will be
    *                  summed. It is set to true by default.
    */
-  SoftMarginLoss(const bool reduction = true);
+  SoftMarginLossType(const bool reduction = true);
 
   /**
    * Computes the Soft Margin Loss function.
@@ -78,7 +78,9 @@ class SoftMarginLoss
  private:
   //! The boolean value that tells if reduction is sum or mean.
   bool reduction;
-}; // class SoftMarginLoss
+}; // class SoftMarginLossType
+
+typedef SoftMarginLossType<arma::mat> SoftMarginLoss;
 
 } // namespace ann
 } // namespace mlpack

src/mlpack/methods/ann/loss_functions/soft_margin_loss_impl.hpp

@@ -19,14 +19,14 @@ namespace mlpack {
 namespace ann /** Artifical Neural Network. */ {
 
 template<typename MatType>
-SoftMarginLoss<MatType>::
-SoftMarginLoss(const bool reduction) : reduction(reduction)
+SoftMarginLossType<MatType>::
+SoftMarginLossType(const bool reduction) : reduction(reduction)
 {
   // Nothing to do here.
 }
 
 template<typename MatType>
-typename MatType::elem_type SoftMarginLoss<MatType>::Forward(
+typename MatType::elem_type SoftMarginLossType<MatType>::Forward(
     const MatType& prediction, const MatType& target)
 {
   MatType loss = arma::log(1 + arma::exp(-target % prediction));
@@ -39,7 +39,7 @@ typename MatType::elem_type SoftMarginLoss<MatType>::Forward(
 }
 
 template<typename MatType>
-void SoftMarginLoss<MatType>::Backward(
+void SoftMarginLossType<MatType>::Backward(
     const MatType& prediction,
     const MatType& target,
     MatType& loss)
@@ -56,7 +56,7 @@ void SoftMarginLoss<MatType>::Backward(
 
 template<typename MatType>
 template<typename Archive>
-void SoftMarginLoss<MatType>::serialize(
+void SoftMarginLossType<MatType>::serialize(
     Archive& ar, const uint32_t /* version */)
 {
   ar(CEREAL_NVP(reduction));

src/mlpack/methods/ann/loss_functions/triplet_margin_loss.hpp

@@ -42,17 +42,17 @@ namespace ann /** Artificial Neural Network. */ {
  *         arma::sp_mat or arma::cube).
  */
 template<typename MatType = arma::mat>
-class TripletMarginLoss
+class TripletMarginLossType
 {
  public:
   /**
-   * Create the TripletMarginLoss object.
+   * Create the TripletMarginLossType object.
    * 
    * @param margin The minimum value by which the distance between 
    *               Anchor and Negative sample exceeds the distance 
    *               between Anchor and Positive sample.
    */
-  TripletMarginLoss(const double margin = 1.0);
+  TripletMarginLossType(const double margin = 1.0);
 
   /**
    * Computes the Triplet Margin Loss function.
@@ -88,7 +88,9 @@ class TripletMarginLoss
  private:
   //! The margin value used in calculating Triplet Margin Loss.
   double margin;
-}; // class TripletLossMargin
+}; // class TripletMarginLoss
+
+typedef TripletMarginLossType<arma::mat> TripletMarginLoss;
 
 } // namespace ann
 } // namespace mlpack

src/mlpack/methods/ann/loss_functions/triplet_margin_loss_impl.hpp

@@ -20,14 +20,14 @@ namespace mlpack {
 namespace ann /** Artifical Neural Network. */ {
 
 template<typename MatType>
-TripletMarginLoss<MatType>::TripletMarginLoss(const double margin) :
+TripletMarginLossType<MatType>::TripletMarginLossType(const double margin) :
     margin(margin)
 {
   // Nothing to do here.
 }
 
 template<typename MatType>
-typename MatType::elem_type TripletMarginLoss<MatType>::Forward(
+typename MatType::elem_type TripletMarginLossType<MatType>::Forward(
     const MatType& prediction,
     const MatType& target)
 {
@@ -41,7 +41,7 @@ typename MatType::elem_type TripletMarginLoss<MatType>::Forward(
 }
 
 template<typename MatType>
-void TripletMarginLoss<MatType>::Backward(
+void TripletMarginLossType<MatType>::Backward(
     const MatType& prediction,
     const MatType& target,
     MatType& loss)
@@ -54,7 +54,7 @@ void TripletMarginLoss<MatType>::Backward(
 
 template<typename MatType>
 template<typename Archive>
-void TripletMarginLoss<MatType>::serialize(
+void TripletMarginLossType<MatType>::serialize(
     Archive& ar,
     const unsigned int /* version */)
 {

src/mlpack/methods/ann/loss_functions/vr_class_reward.hpp

@@ -2,7 +2,7 @@
  * @file methods/ann/loss_functions/vr_class_reward.hpp
  * @author Marcus Edel
  *
- * Definition of the VRClassReward class, which implements the variance
+ * Definition of the VRClassRewardType class, which implements the variance
  * reduced classification reinforcement layer.
  *
  * mlpack is free software; you may redistribute it and/or modify it under the
@@ -30,16 +30,16 @@ namespace ann /** Artificial Neural Network. */ {
  *         arma::sp_mat or arma::cube).
  */
 template<typename MatType = arma::mat>
-class VRClassReward
+class VRClassRewardType
 {
  public:
   /**
-   * Create the VRClassReward object.
+   * Create the VRClassRewardType object.
    *
    * @param scale Parameter used to scale the reward.
    * @param sizeAverage Take the average over all batches.
    */
-  VRClassReward(const double scale = 1, const bool sizeAverage = true);
+  VRClassRewardType(const double scale = 1, const bool sizeAverage = true);
 
   /**
    * Ordinary feed forward pass of a neural network, evaluating the function
@@ -112,7 +112,9 @@ class VRClassReward
 
   //! Locally-stored network modules.
   std::vector<Layer<MatType>*> network;
-}; // class VRClassReward
+}; // class VRClassRewardType
+
+typedef VRClassRewardType<arma::mat> VRClassReward;
 
 } // namespace ann
 } // namespace mlpack

src/mlpack/methods/ann/loss_functions/vr_class_reward_impl.hpp

@@ -2,7 +2,7 @@
  * @file methods/ann/loss_functions/vr_class_reward_impl.hpp
  * @author Marcus Edel
  *
- * Implementation of the VRClassReward class, which implements the variance
+ * Implementation of the VRClassRewardType class, which implements the variance
  * reduced classification reinforcement layer.
  *
  * mlpack is free software; you may redistribute it and/or modify it under the
@@ -20,7 +20,7 @@ namespace mlpack {
 namespace ann /** Artificial Neural Network. */ {
 
 template<typename MatType>
-VRClassReward<MatType>::VRClassReward(
+VRClassRewardType<MatType>::VRClassRewardType(
     const double scale,
     const bool sizeAverage) :
     scale(scale),
@@ -31,7 +31,7 @@ VRClassReward<MatType>::VRClassReward(
 }
 
 template<typename MatType>
-typename MatType::elem_type VRClassReward<MatType>::Forward(
+typename MatType::elem_type VRClassRewardType<MatType>::Forward(
     const MatType& input, const MatType& target)
 {
   double output = 0;
@@ -61,7 +61,7 @@ typename MatType::elem_type VRClassReward<MatType>::Forward(
 }
 
 template<typename MatType>
-void VRClassReward<MatType>::Backward(
+void VRClassRewardType<MatType>::Backward(
     const MatType& input,
     const MatType& target,
     MatType& output)
@@ -89,7 +89,7 @@ void VRClassReward<MatType>::Backward(
 
 template<typename MatType>
 template<typename Archive>
-void VRClassReward<MatType>::serialize(
+void VRClassRewardType<MatType>::serialize(
     Archive& ar, const uint32_t /* version */)
 {
   ar(scale);

src/mlpack/methods/ann/rnn.hpp

@@ -27,7 +27,7 @@ namespace ann /** Artificial Neural Network. */ {
  * @tparam InitializationRuleType Rule used to initialize the weight matrix.
  */
 template<
-    typename OutputLayerType = NegativeLogLikelihood<>,
+    typename OutputLayerType = NegativeLogLikelihood,
     typename InitializationRuleType = RandomInitialization,
     typename MatType = arma::mat>
 class RNN

src/mlpack/tests/ann_layer_test.cpp

@@ -139,7 +139,7 @@ TEST_CASE("GradientAddLayerTest", "[ANNLayerTest]")
         input(arma::randu(10, 1)),
         target(arma::mat("0"))
     {
-      model = new FFN<NegativeLogLikelihood<>, NguyenWidrowInitialization>();
+      model = new FFN<NegativeLogLikelihood, NguyenWidrowInitialization>();
       model->ResetData(input, target);
       model->Add<IdentityLayer>();
       model->Add<Linear>(10, 10);
@@ -161,7 +161,7 @@ TEST_CASE("GradientAddLayerTest", "[ANNLayerTest]")
 
     arma::mat& Parameters() { return model->Parameters(); }
 
-    FFN<NegativeLogLikelihood<>, NguyenWidrowInitialization>* model;
+    FFN<NegativeLogLikelihood, NguyenWidrowInitialization>* model;
     arma::mat input, target;
   } function;
 
@@ -466,7 +466,7 @@ TEST_CASE("NoAlphaDropoutTest", "[ANNLayerTest]")
 //         input(arma::randu(10, 1)),
 //         target(arma::mat("1"))
 //     {
-//       model = new FFN<NegativeLogLikelihood<>, NguyenWidrowInitialization>();
+//       model = new FFN<NegativeLogLikelihood, NguyenWidrowInitialization>();
 //       model->ResetData(input, target);
 //       model->Add<IdentityLayer<> >();
 //       model->Add<Linear<> >(10, 10);
@@ -488,7 +488,7 @@ TEST_CASE("NoAlphaDropoutTest", "[ANNLayerTest]")
 
 //     arma::mat& Parameters() { return model->Parameters(); }
 
-//     FFN<NegativeLogLikelihood<>, NguyenWidrowInitialization>* model;
+//     FFN<NegativeLogLikelihood, NguyenWidrowInitialization>* model;
 //     arma::mat input, target;
 //   } function;
 
@@ -579,7 +579,7 @@ TEST_CASE("GradientLinear3DLayerTest", "[ANNLayerTest]")
       target(1, 1) = 1;
       target(1, 2) = 1;
 
-      model = new FFN<MeanSquaredError<>, RandomInitialization>();
+      model = new FFN<MeanSquaredError, RandomInitialization>();
       model->ResetData(input, target);
       model->Add<Linear3D>(outSize);
       model->InputDimensions() = std::vector<size_t>{ 4, 2 };
@@ -599,7 +599,7 @@ TEST_CASE("GradientLinear3DLayerTest", "[ANNLayerTest]")
 
     arma::mat& Parameters() { return model->Parameters(); }
 
-    FFN<MeanSquaredError<>, RandomInitialization>* model;
+    FFN<MeanSquaredError, RandomInitialization>* model;
     arma::mat input, target;
     const size_t inSize;
     const size_t outSize;
@@ -655,7 +655,7 @@ TEST_CASE("GradientLinear3DLayerTest", "[ANNLayerTest]")
 //         input(arma::randu(10, 1)),
 //         target(arma::mat("1"))
 //     {
-//       model = new FFN<NegativeLogLikelihood<>, NguyenWidrowInitialization>();
+//       model = new FFN<NegativeLogLikelihood, NguyenWidrowInitialization>();
 //       model->ResetData(input, target);
 //       model->Add<IdentityLayer<> >();
 //       model->Add<NoisyLinear<> >(10, 10);
@@ -677,7 +677,7 @@ TEST_CASE("GradientLinear3DLayerTest", "[ANNLayerTest]")
 
 //     arma::mat& Parameters() { return model->Parameters(); }
 
-//     FFN<NegativeLogLikelihood<>, NguyenWidrowInitialization>* model;
+//     FFN<NegativeLogLikelihood, NguyenWidrowInitialization>* model;
 //     arma::mat input, target;
 //   } function;
 
@@ -808,7 +808,7 @@ TEST_CASE("GradientLinearNoBiasLayerTest", "[ANNLayerTest]")
         input(arma::randu(10, 1)),
         target(arma::mat("0"))
     {
-      model = new FFN<NegativeLogLikelihood<>, NguyenWidrowInitialization>();
+      model = new FFN<NegativeLogLikelihood, NguyenWidrowInitialization>();
       model->ResetData(input, target);
       model->Add<Linear>(10);
       model->Add<LinearNoBias>(2);
@@ -829,7 +829,7 @@ TEST_CASE("GradientLinearNoBiasLayerTest", "[ANNLayerTest]")
 
     arma::mat& Parameters() { return model->Parameters(); }
 
-    FFN<NegativeLogLikelihood<>, NguyenWidrowInitialization>* model;
+    FFN<NegativeLogLikelihood, NguyenWidrowInitialization>* model;
     arma::mat input, target;
   } function;
 
@@ -843,7 +843,7 @@ TEST_CASE("GradientLinearNoBiasLayerTest", "[ANNLayerTest]")
 // {
 //   for (size_t i = 0; i < 5; ++i)
 //   {
-//     NegativeLogLikelihood<> module;
+//     NegativeLogLikelihood module;
 //     const size_t inputElements = math::RandInt(5, 100);
 //     arma::mat input;
 //     RandomInitialization init(0, 1);
@@ -909,8 +909,8 @@ TEST_CASE("GradientFlexibleReLULayerTest", "[ANNLayerTest]")
         input(arma::randu(2, 1)),
         target(arma::mat("0"))
     {
-      model = new FFN<NegativeLogLikelihood<>, RandomInitialization>(
-          NegativeLogLikelihood<>(), RandomInitialization(0.1, 0.5));
+      model = new FFN<NegativeLogLikelihood, RandomInitialization>(
+          NegativeLogLikelihood(), RandomInitialization(0.1, 0.5));
 
       model->ResetData(input, target);
       model->Add<Linear>(2, 2);
@@ -933,7 +933,7 @@ TEST_CASE("GradientFlexibleReLULayerTest", "[ANNLayerTest]")
 
     arma::mat& Parameters() { return model->Parameters(); }
 
-    FFN<NegativeLogLikelihood<>, RandomInitialization>* model;
+    FFN<NegativeLogLikelihood, RandomInitialization>* model;
     arma::mat input, target;
   } function;
 
@@ -1129,8 +1129,8 @@ TEST_CASE("SimpleJoinLayerTest", "[ANNLayerTest]")
 //   RandomInitialization init(0.5, 0.5);
 
 //   // Create model with user defined rho parameter.
-//   RNN<NegativeLogLikelihood<>, RandomInitialization> modelA(
-//       rho, false, NegativeLogLikelihood<>(), init);
+//   RNN<NegativeLogLikelihood, RandomInitialization> modelA(
+//       rho, false, NegativeLogLikelihood(), init);
 //   modelA.Add<IdentityLayer<> >();
 //   modelA.Add<Linear<> >(1, 10);
 
@@ -1139,8 +1139,8 @@ TEST_CASE("SimpleJoinLayerTest", "[ANNLayerTest]")
 //   modelA.Add<LogSoftMax<> >();
 
 //   // Create model without user defined rho parameter.
-//   RNN<NegativeLogLikelihood<> > modelB(
-//       rho, false, NegativeLogLikelihood<>(), init);
+//   RNN<NegativeLogLikelihood> modelB(
+//       rho, false, NegativeLogLikelihood(), init);
 //   modelB.Add<IdentityLayer<> >();
 //   modelB.Add<Linear<> >(1, 10);
 
@@ -1169,7 +1169,7 @@ TEST_CASE("SimpleJoinLayerTest", "[ANNLayerTest]")
 //     {
 //       const size_t rho = 5;
 
-//       model = new RNN<NegativeLogLikelihood<> >(rho);
+//       model = new RNN<NegativeLogLikelihood>(rho);
 //       model->ResetData(input, target);
 //       model->Add<IdentityLayer<> >();
 //       model->Add<Linear<> >(1, 10);
@@ -1191,7 +1191,7 @@ TEST_CASE("SimpleJoinLayerTest", "[ANNLayerTest]")
 
 //     arma::mat& Parameters() { return model->Parameters(); }
 
-//     RNN<NegativeLogLikelihood<> >* model;
+//     RNN<NegativeLogLikelihood>* model;
 //     arma::cube input, target;
 //   } function;
 
@@ -1233,8 +1233,8 @@ TEST_CASE("SimpleJoinLayerTest", "[ANNLayerTest]")
 //   RandomInitialization init(0.5, 0.5);
 
 //   // Create model with user defined rho parameter.
-//   RNN<NegativeLogLikelihood<>, RandomInitialization> modelA(
-//       rho, false, NegativeLogLikelihood<>(), init);
+//   RNN<NegativeLogLikelihood, RandomInitialization> modelA(
+//       rho, false, NegativeLogLikelihood(), init);
 //   modelA.Add<IdentityLayer<> >();
 //   modelA.Add<Linear<> >(1, 10);
 
@@ -1243,8 +1243,8 @@ TEST_CASE("SimpleJoinLayerTest", "[ANNLayerTest]")
 //   modelA.Add<LogSoftMax<> >();
 
 //   // Create model without user defined rho parameter.
-//   RNN<NegativeLogLikelihood<> > modelB(
-//       rho, false, NegativeLogLikelihood<>(), init);
+//   RNN<NegativeLogLikelihood> modelB(
+//       rho, false, NegativeLogLikelihood(), init);
 //   modelB.Add<IdentityLayer<> >();
 //   modelB.Add<Linear<> >(1, 10);
 
@@ -1273,7 +1273,7 @@ TEST_CASE("SimpleJoinLayerTest", "[ANNLayerTest]")
 //     {
 //       const size_t rho = 5;
 
-//       model = new RNN<NegativeLogLikelihood<> >(rho);
+//       model = new RNN<NegativeLogLikelihood>(rho);
 //       model->ResetData(input, target);
 //       model->Add<IdentityLayer<> >();
 //       model->Add<Linear<> >(1, 10);
@@ -1295,7 +1295,7 @@ TEST_CASE("SimpleJoinLayerTest", "[ANNLayerTest]")
 
 //     arma::mat& Parameters() { return model->Parameters(); }
 
-//     RNN<NegativeLogLikelihood<> >* model;
+//     RNN<NegativeLogLikelihood>* model;
 //     arma::cube input, target;
 //   } function;
 
@@ -1338,16 +1338,16 @@ TEST_CASE("SimpleJoinLayerTest", "[ANNLayerTest]")
 //   arma::cube target = arma::ones(1, 1, 5);
 //   const size_t rho = 5;
 
-//   RNN<NegativeLogLikelihood<> > *model1 =
-//       new RNN<NegativeLogLikelihood<> >(rho);
+//   RNN<NegativeLogLikelihood> *model1 =
+//       new RNN<NegativeLogLikelihood>(rho);
 //   model1->ResetData(input, target);
 //   model1->Add<IdentityLayer<> >();
 //   model1->Add<Linear<> >(1, 10);
 //   model1->Add<FastLSTM<> >(10, 3, rho);
 //   model1->Add<LogSoftMax<> >();
 
-//   RNN<NegativeLogLikelihood<> > *model2 =
-//      new RNN<NegativeLogLikelihood<> >(rho);
+//   RNN<NegativeLogLikelihood> *model2 =
+//      new RNN<NegativeLogLikelihood>(rho);
 //   model2->ResetData(input, target);
 //   model2->Add<IdentityLayer<> >();
 //   model2->Add<Linear<> >(1, 10);
@@ -1370,16 +1370,16 @@ TEST_CASE("SimpleJoinLayerTest", "[ANNLayerTest]")
 //   arma::cube target = arma::ones(1, 1, 5);
 //   const size_t rho = 5;
 
-//   RNN<NegativeLogLikelihood<> > *model1 =
-//       new RNN<NegativeLogLikelihood<> >(rho);
+//   RNN<NegativeLogLikelihood> *model1 =
+//       new RNN<NegativeLogLikelihood>(rho);
 //   model1->ResetData(input, target);
 //   model1->Add<IdentityLayer<> >();
 //   model1->Add<Linear<> >(1, 10);
 //   model1->Add<LSTM<> >(10, 3, rho);
 //   model1->Add<LogSoftMax<> >();
 
-//   RNN<NegativeLogLikelihood<> > *model2 =
-//      new RNN<NegativeLogLikelihood<> >(rho);
+//   RNN<NegativeLogLikelihood> *model2 =
+//      new RNN<NegativeLogLikelihood>(rho);
 //   model2->ResetData(input, target);
 //   model2->Add<IdentityLayer<> >();
 //   model2->Add<Linear<> >(1, 10);
@@ -1605,7 +1605,7 @@ TEST_CASE("SimpleJoinLayerTest", "[ANNLayerTest]")
 //     {
 //       const size_t rho = 5;
 
-//       model = new RNN<NegativeLogLikelihood<> >(rho);
+//       model = new RNN<NegativeLogLikelihood>(rho);
 //       model->ResetData(input, target);
 //       model->Add<IdentityLayer<> >();
 //       model->Add<Linear<> >(1, 10);
@@ -1628,7 +1628,7 @@ TEST_CASE("SimpleJoinLayerTest", "[ANNLayerTest]")
 
 //     arma::mat& Parameters() { return model->Parameters(); }
 
-//     RNN<NegativeLogLikelihood<> >* model;
+//     RNN<NegativeLogLikelihood>* model;
 //     arma::cube input, target;
 //   } function;
 
@@ -1734,8 +1734,8 @@ TEST_CASE("SimpleJoinLayerTest", "[ANNLayerTest]")
 //   RandomInitialization init(0.5, 0.5);
 //
 //   // Create model with user defined rho parameter.
-//   RNN<NegativeLogLikelihood<>, RandomInitialization> modelA(
-//       rho, false, NegativeLogLikelihood<>(), init);
+//   RNN<NegativeLogLikelihood, RandomInitialization> modelA(
+//       rho, false, NegativeLogLikelihood(), init);
 //   modelA.Add<IdentityLayer<> >();
 //   modelA.Add<Linear<> >(1, 10);
 //
@@ -1744,8 +1744,8 @@ TEST_CASE("SimpleJoinLayerTest", "[ANNLayerTest]")
 //   modelA.Add<LogSoftMax<> >();
 //
 //   // Create model without user defined rho parameter.
-//   RNN<NegativeLogLikelihood<> > modelB(
-//       rho, false, NegativeLogLikelihood<>(), init);
+//   RNN<NegativeLogLikelihood> modelB(
+//       rho, false, NegativeLogLikelihood(), init);
 //   modelB.Add<IdentityLayer<> >();
 //   modelB.Add<Linear<> >(1, 10);
 //
@@ -1774,7 +1774,7 @@ TEST_CASE("SimpleJoinLayerTest", "[ANNLayerTest]")
 //     {
 //       const size_t rho = 5;
 //
-//       model = new RNN<NegativeLogLikelihood<> >(rho);
+//       model = new RNN<NegativeLogLikelihood>(rho);
 //       model->ResetData(input, target);
 //       model->Add<IdentityLayer<> >();
 //       model->Add<Linear<> >(1, 10);
@@ -1796,7 +1796,7 @@ TEST_CASE("SimpleJoinLayerTest", "[ANNLayerTest]")
 //
 //     arma::mat& Parameters() { return model->Parameters(); }
 //
-//     RNN<NegativeLogLikelihood<> >* model;
+//     RNN<NegativeLogLikelihood>* model;
 //     arma::cube input, target;
 //   } function;
 //
@@ -1838,8 +1838,8 @@ TEST_CASE("SimpleJoinLayerTest", "[ANNLayerTest]")
 //   RandomInitialization init(0.5, 0.5);
 //
 //   // Create model with user defined rho parameter.
-//   RNN<NegativeLogLikelihood<>, RandomInitialization> modelA(
-//       rho, false, NegativeLogLikelihood<>(), init);
+//   RNN<NegativeLogLikelihood, RandomInitialization> modelA(
+//       rho, false, NegativeLogLikelihood(), init);
 //   modelA.Add<IdentityLayer<> >();
 //   modelA.Add<Linear<> >(1, 10);
 //
@@ -1848,8 +1848,8 @@ TEST_CASE("SimpleJoinLayerTest", "[ANNLayerTest]")
 //   modelA.Add<LogSoftMax<> >();
 //
 //   // Create model without user defined rho parameter.
-//   RNN<NegativeLogLikelihood<> > modelB(
-//       rho, false, NegativeLogLikelihood<>(), init);
+//   RNN<NegativeLogLikelihood> modelB(
+//       rho, false, NegativeLogLikelihood(), init);
 //   modelB.Add<IdentityLayer<> >();
 //   modelB.Add<Linear<> >(1, 10);
 //
@@ -1878,7 +1878,7 @@ TEST_CASE("SimpleJoinLayerTest", "[ANNLayerTest]")
 //     {
 //       const size_t rho = 5;
 //
-//       model = new RNN<NegativeLogLikelihood<> >(rho);
+//       model = new RNN<NegativeLogLikelihood>(rho);
 //       model->ResetData(input, target);
 //       model->Add<IdentityLayer<> >();
 //       model->Add<Linear<> >(1, 10);
@@ -1900,7 +1900,7 @@ TEST_CASE("SimpleJoinLayerTest", "[ANNLayerTest]")
 //
 //     arma::mat& Parameters() { return model->Parameters(); }
 //
-//     RNN<NegativeLogLikelihood<> >* model;
+//     RNN<NegativeLogLikelihood>* model;
 //     arma::cube input, target;
 //   } function;
 //
@@ -1943,16 +1943,16 @@ TEST_CASE("SimpleJoinLayerTest", "[ANNLayerTest]")
 //   arma::cube target = arma::ones(1, 1, 5);
 //   const size_t rho = 5;
 //
-//   RNN<NegativeLogLikelihood<> > *model1 =
-//       new RNN<NegativeLogLikelihood<> >(rho);
+//   RNN<NegativeLogLikelihood> *model1 =
+//       new RNN<NegativeLogLikelihood>(rho);
 //   model1->ResetData(input, target);
 //   model1->Add<IdentityLayer<> >();
 //   model1->Add<Linear<> >(1, 10);
 //   model1->Add<FastLSTM<> >(10, 3, rho);
 //   model1->Add<LogSoftMax<> >();
 //
-//   RNN<NegativeLogLikelihood<> > *model2 =
-//      new RNN<NegativeLogLikelihood<> >(rho);
+//   RNN<NegativeLogLikelihood> *model2 =
+//      new RNN<NegativeLogLikelihood>(rho);
 //   model2->ResetData(input, target);
 //   model2->Add<IdentityLayer<> >();
 //   model2->Add<Linear<> >(1, 10);
@@ -1975,16 +1975,16 @@ TEST_CASE("SimpleJoinLayerTest", "[ANNLayerTest]")
 //   arma::cube target = arma::ones(1, 1, 5);
 //   const size_t rho = 5;
 //
-//   RNN<NegativeLogLikelihood<> > *model1 =
-//       new RNN<NegativeLogLikelihood<> >(rho);
+//   RNN<NegativeLogLikelihood> *model1 =
+//       new RNN<NegativeLogLikelihood>(rho);
 //   model1->ResetData(input, target);
 //   model1->Add<IdentityLayer<> >();
 //   model1->Add<Linear<> >(1, 10);
 //   model1->Add<LSTM<> >(10, 3, rho);
 //   model1->Add<LogSoftMax<> >();
 //
-//   RNN<NegativeLogLikelihood<> > *model2 =
-//      new RNN<NegativeLogLikelihood<> >(rho);
+//   RNN<NegativeLogLikelihood> *model2 =
+//      new RNN<NegativeLogLikelihood>(rho);
 //   model2->ResetData(input, target);
 //   model2->Add<IdentityLayer<> >();
 //   model2->Add<Linear<> >(1, 10);
@@ -2210,7 +2210,7 @@ TEST_CASE("SimpleJoinLayerTest", "[ANNLayerTest]")
 //     {
 //       const size_t rho = 5;
 //
-//       model = new RNN<NegativeLogLikelihood<> >(rho);
+//       model = new RNN<NegativeLogLikelihood>(rho);
 //       model->ResetData(input, target);
 //       model->Add<IdentityLayer<> >();
 //       model->Add<Linear<> >(1, 10);
@@ -2233,7 +2233,7 @@ TEST_CASE("SimpleJoinLayerTest", "[ANNLayerTest]")
 //
 //     arma::mat& Parameters() { return model->Parameters(); }
 //
-//     RNN<NegativeLogLikelihood<> >* model;
+//     RNN<NegativeLogLikelihood>* model;
 //     arma::cube input, target;
 //   } function;
 //
@@ -2449,7 +2449,7 @@ TEST_CASE("ConcatLayerParametersTest", "[ANNLayerTest]")
 //         input(arma::randu(10, 1)),
 //         target(arma::mat("0"))
 //     {
-//       model = new FFN<NegativeLogLikelihood<>, NguyenWidrowInitialization>();
+//       model = new FFN<NegativeLogLikelihood, NguyenWidrowInitialization>();
 //       model->ResetData(input, target);
 //       model->Add<IdentityLayer>();
 //       model->Add<Linear>(10, 10);
@@ -2475,7 +2475,7 @@ TEST_CASE("ConcatLayerParametersTest", "[ANNLayerTest]")
 
 //     arma::mat& Parameters() { return model->Parameters(); }
 
-//     FFN<NegativeLogLikelihood<>, NguyenWidrowInitialization>* model;
+//     FFN<NegativeLogLikelihood, NguyenWidrowInitialization>* model;
 //     Concat* concat;
 //     arma::mat input, target;
 //   } function;
@@ -2517,7 +2517,7 @@ TEST_CASE("GradientConcatenateLayerTest", "[ANNLayerTest]")
         input(arma::randu(10, 1)),
         target(arma::mat("0"))
     {
-      model = new FFN<NegativeLogLikelihood<>, NguyenWidrowInitialization>();
+      model = new FFN<NegativeLogLikelihood, NguyenWidrowInitialization>();
       model->ResetData(input, target);
       model->Add<IdentityLayer>();
       model->Add<Linear>(10, 5);
@@ -2546,7 +2546,7 @@ TEST_CASE("GradientConcatenateLayerTest", "[ANNLayerTest]")
 
     arma::mat& Parameters() { return model->Parameters(); }
 
-    FFN<NegativeLogLikelihood<>, NguyenWidrowInitialization>* model;
+    FFN<NegativeLogLikelihood, NguyenWidrowInitialization>* model;
     Concatenate* concatenate;
     arma::mat input, target;
   } function;
@@ -2725,7 +2725,7 @@ TEST_CASE("GradientSoftmaxTest", "[ANNLayerTest]")
         input(arma::randu(10, 1)),
         target(arma::mat("1; 0"))
     {
-      model = new FFN<MeanSquaredError<>, RandomInitialization>;
+      model = new FFN<MeanSquaredError, RandomInitialization>;
       model->ResetData(input, target);
       model->Add<Linear>(10, 10);
       model->Add<ReLULayer>();
@@ -2747,7 +2747,7 @@ TEST_CASE("GradientSoftmaxTest", "[ANNLayerTest]")
 
     arma::mat& Parameters() { return model->Parameters(); }
 
-    FFN<MeanSquaredError<> >* model;
+    FFN<MeanSquaredError>* model;
     arma::mat input, target;
   } function;
 
@@ -3086,7 +3086,7 @@ TEST_CASE("SimpleBicubicInterpolationLayerTest", "[ANNLayerTest]")
 //           input(arma::randn(32, 2048)),
 //           target(arma::zeros(1, 2048))
 //       {
-//         model = new FFN<NegativeLogLikelihood<>, NguyenWidrowInitialization>();
+//         model = new FFN<NegativeLogLikelihood, NguyenWidrowInitialization>();
 //         model->ResetData(input, target);
 //         model->Add<IdentityLayer<> >();
 //         model->Add<Linear<> >(32, 4);
@@ -3109,7 +3109,7 @@ TEST_CASE("SimpleBicubicInterpolationLayerTest", "[ANNLayerTest]")
 
 //       arma::mat& Parameters() { return model->Parameters(); }
 
-//       FFN<NegativeLogLikelihood<>, NguyenWidrowInitialization>* model;
+//       FFN<NegativeLogLikelihood, NguyenWidrowInitialization>* model;
 //       arma::mat input, target;
 //     } function;
 
@@ -3160,7 +3160,7 @@ TEST_CASE("GradientVirtualBatchNormTest", "[ANNLayerTest]")
     {
       arma::mat referenceBatch = arma::mat(input.memptr(), input.n_rows, 4);
 
-      model = new FFN<NegativeLogLikelihood<>, NguyenWidrowInitialization>();
+      model = new FFN<NegativeLogLikelihood, NguyenWidrowInitialization>();
       model->ResetData(input, target);
       model->Add<IdentityLayer>();
       model->Add<Linear>(5, 5);
@@ -3183,7 +3183,7 @@ TEST_CASE("GradientVirtualBatchNormTest", "[ANNLayerTest]")
 
     arma::mat& Parameters() { return model->Parameters(); }
 
-    FFN<NegativeLogLikelihood<>, NguyenWidrowInitialization>* model;
+    FFN<NegativeLogLikelihood, NguyenWidrowInitialization>* model;
     arma::mat input, target;
   } function;
 
@@ -3221,7 +3221,7 @@ TEST_CASE("VirtualBatchNormLayerParametersTest", "[ANNLayerTest]")
 //         input(arma::randn(5, 4)),
 //         target(arma::zeros(1, 4))
 //     {
-//       model = new FFN<NegativeLogLikelihood<>, NguyenWidrowInitialization>();
+//       model = new FFN<NegativeLogLikelihood, NguyenWidrowInitialization>();
 //       model->ResetData(input, target);
 //       model->Add<IdentityLayer<> >();
 //       model->Add<Linear<> >(5, 5);
@@ -3242,7 +3242,7 @@ TEST_CASE("VirtualBatchNormLayerParametersTest", "[ANNLayerTest]")
 
 //     arma::mat& Parameters() { return model->Parameters(); }
 
-//     FFN<NegativeLogLikelihood<>, NguyenWidrowInitialization>* model;
+//     FFN<NegativeLogLikelihood, NguyenWidrowInitialization>* model;
 //     arma::mat input, target;
 //   } function;
 
@@ -3399,7 +3399,7 @@ TEST_CASE("GradientTransposedConvolutionLayerTest", "[ANNLayerTest]")
           input(arma::linspace<arma::colvec>(0, 35, 36)),
           target(arma::mat("0"))
       {
-        model = new FFN<NegativeLogLikelihood<>, RandomInitialization>();
+        model = new FFN<NegativeLogLikelihood, RandomInitialization>();
         model->ResetData(input, target);
         model->Add<TransposedConvolution>(1, 1, 3, 3, 2, 2, 1, 1, 6, 6, 12, 12);
         model->Add<LogSoftMax>();
@@ -3419,7 +3419,7 @@ TEST_CASE("GradientTransposedConvolutionLayerTest", "[ANNLayerTest]")
 
       arma::mat& Parameters() { return model->Parameters(); }
 
-      FFN<NegativeLogLikelihood<>, RandomInitialization>* model;
+      FFN<NegativeLogLikelihood, RandomInitialization>* model;
       arma::mat input, target;
     } function;
 
@@ -3565,7 +3565,7 @@ TEST_CASE("SimpleMultiplyMergeLayerTest", "[ANNLayerTest]")
 //         input(arma::linspace<arma::colvec>(0, 35, 36)),
 //         target(arma::mat("0"))
 //     {
-//       model = new FFN<NegativeLogLikelihood<>, RandomInitialization>();
+//       model = new FFN<NegativeLogLikelihood, RandomInitialization>();
 //       model->ResetData(input, target);
 //       model->Add<IdentityLayer<> >();
 //       model->Add<AtrousConvolution<> >(1, 1, 3, 3, 1, 1, 0, 0, 6, 6, 2, 2);
@@ -3586,7 +3586,7 @@ TEST_CASE("SimpleMultiplyMergeLayerTest", "[ANNLayerTest]")
 
 //     arma::mat& Parameters() { return model->Parameters(); }
 
-//     FFN<NegativeLogLikelihood<>, RandomInitialization>* model;
+//     FFN<NegativeLogLikelihood, RandomInitialization>* model;
 //     arma::mat input, target;
 //   } function;
 
@@ -3747,7 +3747,7 @@ TEST_CASE("SimpleMultiplyMergeLayerTest", "[ANNLayerTest]")
 //         input(arma::randn(10, 256)),
 //         target(arma::zeros(1, 256))
 //     {
-//       model = new FFN<NegativeLogLikelihood<>, NguyenWidrowInitialization>();
+//       model = new FFN<NegativeLogLikelihood, NguyenWidrowInitialization>();
 //       model->ResetData(input, target);
 //       model->Add<IdentityLayer<> >();
 //       model->Add<Linear<> >(10, 10);
@@ -3770,7 +3770,7 @@ TEST_CASE("SimpleMultiplyMergeLayerTest", "[ANNLayerTest]")
 //
 //     arma::mat& Parameters() { return model->Parameters(); }
 //
-//     FFN<NegativeLogLikelihood<>, NguyenWidrowInitialization>* model;
+//     FFN<NegativeLogLikelihood, NguyenWidrowInitialization>* model;
 //     arma::mat input, target;
 //   } function;
 //
@@ -3824,7 +3824,7 @@ TEST_CASE("GradientLayerNormTest", "[ANNLayerTest]")
         input(arma::randn(10, 256)),
         target(arma::zeros(1, 256))
     {
-      model = new FFN<NegativeLogLikelihood<>, NguyenWidrowInitialization>();
+      model = new FFN<NegativeLogLikelihood, NguyenWidrowInitialization>();
       model->ResetData(input, target);
       model->Add<IdentityLayer>();
       model->Add<Linear>(10, 10);
@@ -3847,7 +3847,7 @@ TEST_CASE("GradientLayerNormTest", "[ANNLayerTest]")
 
     arma::mat& Parameters() { return model->Parameters(); }
 
-    FFN<NegativeLogLikelihood<>, NguyenWidrowInitialization>* model;
+    FFN<NegativeLogLikelihood, NguyenWidrowInitialization>* model;
     arma::mat input, target;
   } function;
 
@@ -4154,7 +4154,7 @@ TEST_CASE("GradientReparametrizationLayerTest", "[ANNLayerTest]")
         input(arma::randu(10, 1)),
         target(arma::mat("0"))
     {
-      model = new FFN<NegativeLogLikelihood<>, NguyenWidrowInitialization>();
+      model = new FFN<NegativeLogLikelihood, NguyenWidrowInitialization>();
       model->ResetData(input, target);
       model->Add<IdentityLayer>();
       model->Add<Linear>(10, 6);
@@ -4177,7 +4177,7 @@ TEST_CASE("GradientReparametrizationLayerTest", "[ANNLayerTest]")
 
     arma::mat& Parameters() { return model->Parameters(); }
 
-    FFN<NegativeLogLikelihood<>, NguyenWidrowInitialization>* model;
+    FFN<NegativeLogLikelihood, NguyenWidrowInitialization>* model;
     arma::mat input, target;
   } function;
 
@@ -4197,7 +4197,7 @@ TEST_CASE("GradientReparametrizationLayerBetaTest", "[ANNLayerTest]")
         input(arma::randu(10, 2)),
         target(arma::mat("0 0"))
     {
-      model = new FFN<NegativeLogLikelihood<>, NguyenWidrowInitialization>();
+      model = new FFN<NegativeLogLikelihood, NguyenWidrowInitialization>();
       model->ResetData(input, target);
       model->Add<IdentityLayer>();
       model->Add<Linear>(10, 6);
@@ -4221,7 +4221,7 @@ TEST_CASE("GradientReparametrizationLayerBetaTest", "[ANNLayerTest]")
 
     arma::mat& Parameters() { return model->Parameters(); }
 
-    FFN<NegativeLogLikelihood<>, NguyenWidrowInitialization>* model;
+    FFN<NegativeLogLikelihood, NguyenWidrowInitialization>* model;
     arma::mat input, target;
   } function;
 
@@ -4356,7 +4356,7 @@ TEST_CASE("HighwayLayerParametersTest", "[ANNLayerTest]")
 //         input(arma::randu(5, 1)),
 //         target(arma::mat("0"))
 //     {
-//       model = new FFN<NegativeLogLikelihood<>, NguyenWidrowInitialization>();
+//       model = new FFN<NegativeLogLikelihood, NguyenWidrowInitialization>();
 //       model->ResetData(input, target);
 //       model->Add<IdentityLayer>();
 //       model->Add<Linear>(5, 10);
@@ -4386,7 +4386,7 @@ TEST_CASE("HighwayLayerParametersTest", "[ANNLayerTest]")
 
 //     arma::mat& Parameters() { return model->Parameters(); }
 
-//     FFN<NegativeLogLikelihood<>, NguyenWidrowInitialization>* model;
+//     FFN<NegativeLogLikelihood, NguyenWidrowInitialization>* model;
 //     Highway* highway;
 //     arma::mat input, target;
 //   } function;
@@ -4406,7 +4406,7 @@ TEST_CASE("HighwayLayerParametersTest", "[ANNLayerTest]")
 //         input(arma::randu(10, 1)),
 //         target(arma::mat("0"))
 //     {
-//       model = new FFN<NegativeLogLikelihood<>, NguyenWidrowInitialization>();
+//       model = new FFN<NegativeLogLikelihood, NguyenWidrowInitialization>();
 //       model->ResetData(input, target);
 //       model->Add<IdentityLayer>();
 //       model->Add<Linear>(10, 10);
@@ -4435,7 +4435,7 @@ TEST_CASE("HighwayLayerParametersTest", "[ANNLayerTest]")
 
 //     arma::mat& Parameters() { return model->Parameters(); }
 
-//     FFN<NegativeLogLikelihood<>, NguyenWidrowInitialization>* model;
+//     FFN<NegativeLogLikelihood, NguyenWidrowInitialization>* model;
 //     Sequential* sequential;
 //     arma::mat input, target;
 //   } function;
@@ -4455,7 +4455,7 @@ TEST_CASE("HighwayLayerParametersTest", "[ANNLayerTest]")
 //         input(arma::randu(10, 1)),
 //         target(arma::mat("0"))
 //     {
-//       model = new FFN<NegativeLogLikelihood<>, NguyenWidrowInitialization>();
+//       model = new FFN<NegativeLogLikelihood, NguyenWidrowInitialization>();
 //       model->ResetData(input, target);
 //       model->Add<Linear>(10, 10);
 
@@ -4480,7 +4480,7 @@ TEST_CASE("HighwayLayerParametersTest", "[ANNLayerTest]")
 
 //     arma::mat& Parameters() { return model->Parameters(); }
 
-//     FFN<NegativeLogLikelihood<>, NguyenWidrowInitialization>* model;
+//     FFN<NegativeLogLikelihood, NguyenWidrowInitialization>* model;
 //     WeightNorm* weightNorm;
 //     arma::mat input, target;
 //   } function;
@@ -4521,7 +4521,7 @@ TEST_CASE("HighwayLayerParametersTest", "[ANNLayerTest]")
 //   arma::mat input(5, 100, arma::fill::randu);
 //   arma::mat output(5, 100, arma::fill::randu);
 
-//   FFN<NegativeLogLikelihood<>, ann::RandomInitialization> model;
+//   FFN<NegativeLogLikelihood, ann::RandomInitialization> model;
 //   model.Add<Linear<>>(input.n_rows, 10);
 //   model.Add<LayerType>(layer);
 //   model.Add<ReLULayer<>>();
@@ -4535,7 +4535,7 @@ TEST_CASE("HighwayLayerParametersTest", "[ANNLayerTest]")
 //   model.Predict(input, originalOutput);
 
 //   // Now serialize the model.
-//   FFN<NegativeLogLikelihood<>, ann::RandomInitialization> xmlModel, jsonModel,
+//   FFN<NegativeLogLikelihood, ann::RandomInitialization> xmlModel, jsonModel,
 //       binaryModel;
 //   SerializeObjectAll(model, xmlModel, jsonModel, binaryModel);
 
@@ -4678,7 +4678,7 @@ TEST_CASE("GradientConvolutionLayerTest", "[ANNLayerTest]")
         input(arma::linspace<arma::colvec>(0, 35, 36)),
         target(arma::mat("1"))
     {
-      model = new FFN<NegativeLogLikelihood<>, RandomInitialization>();
+      model = new FFN<NegativeLogLikelihood, RandomInitialization>();
       model->ResetData(input, target);
       model->Add<Convolution>(1, 3, 3, 1, 1, std::tuple<size_t, size_t>(0, 0),
           std::tuple<size_t, size_t>(0, 0), "same");
@@ -4701,7 +4701,7 @@ TEST_CASE("GradientConvolutionLayerTest", "[ANNLayerTest]")
 
     arma::mat& Parameters() { return model->Parameters(); }
 
-    FFN<NegativeLogLikelihood<>, RandomInitialization>* model;
+    FFN<NegativeLogLikelihood, RandomInitialization>* model;
     arma::mat input, target;
   } function;
 
@@ -5425,7 +5425,7 @@ TEST_CASE("TransposedConvolutionalLayerOptionalParameterTest", "[ANNLayerTest]")
 //           input(arma::randn(16, 1024)),
 //           target(arma::zeros(1, 1024))
 //       {
-//         model = new FFN<NegativeLogLikelihood<>, NguyenWidrowInitialization>();
+//         model = new FFN<NegativeLogLikelihood, NguyenWidrowInitialization>();
 //         model->ResetData(input, target);
 //         model->Add<IdentityLayer<>>();
 //         model->Add<Convolution<>>(1, 2, 3, 3, 1, 1, 0, 0, 4, 4);
@@ -5448,7 +5448,7 @@ TEST_CASE("TransposedConvolutionalLayerOptionalParameterTest", "[ANNLayerTest]")
 
 //       arma::mat& Parameters() { return model->Parameters(); }
 
-//       FFN<NegativeLogLikelihood<>, NguyenWidrowInitialization>* model;
+//       FFN<NegativeLogLikelihood, NguyenWidrowInitialization>* model;
 //       arma::mat input, target;
 //     } function;
 
@@ -6091,7 +6091,7 @@ TEST_CASE("GradientMultiheadAttentionTest", "[ANNLayerTest]")
       keyPaddingMask = arma::zeros(1, srcSeqLen);
       keyPaddingMask(srcSeqLen - 1) = std::numeric_limits<double>::lowest();
 
-      model = new FFN<NegativeLogLikelihood<>, XavierInitialization>();
+      model = new FFN<NegativeLogLikelihood, XavierInitialization>();
       model->ResetData(input, target);
       // attnModule = new MultiheadAttention(tgtSeqLen, srcSeqLen, embedDim,
       //     nHeads);
@@ -6118,7 +6118,7 @@ TEST_CASE("GradientMultiheadAttentionTest", "[ANNLayerTest]")
 
     arma::mat& Parameters() { return model->Parameters(); }
 
-    FFN<NegativeLogLikelihood<>, XavierInitialization>* model;
+    FFN<NegativeLogLikelihood, XavierInitialization>* model;
     // MultiheadAttention* attnModule;
 
     arma::mat input, target, attnMask, keyPaddingMask;
@@ -6294,7 +6294,7 @@ TEST_CASE("GradientInstanceNormLayerTest", "[ANNLayerTest]")
         arma::mat target;
         target.ones(1, 1024);
 
-        model = new FFN<NegativeLogLikelihood<>, NguyenWidrowInitialization>();
+        model = new FFN<NegativeLogLikelihood, NguyenWidrowInitialization>();
         model->ResetData(input, target);
         model->Add<IdentityLayer<> >();
         model->Add<Convolution<> >(1, 2, 3, 3, 1, 1, 0, 0, 4, 4);
@@ -6317,7 +6317,7 @@ TEST_CASE("GradientInstanceNormLayerTest", "[ANNLayerTest]")
 
       arma::mat& Parameters() { return model->Parameters(); }
 
-      FFN<NegativeLogLikelihood<>, NguyenWidrowInitialization>* model;
+      FFN<NegativeLogLikelihood, NguyenWidrowInitialization>* model;
       arma::mat input, target;
     } function;
 

src/mlpack/tests/callback_test.cpp

@@ -48,7 +48,7 @@ TEST_CASE("FFNCallbackTest", "[CallbackTest]")
   if (!data::Load("lab3.csv", labels))
     FAIL("Cannot load test dataset lab3.csv!");
 
-  FFN<MeanSquaredError<>, RandomInitialization> model;
+  FFN<MeanSquaredError, RandomInitialization> model;
 
   model.Add<Linear>(2);
   model.Add<Sigmoid>();
@@ -74,7 +74,7 @@ TEST_CASE("FFNWithOptimizerCallbackTest", "[CallbackTest]")
   if (!data::Load("lab3.csv", labels))
     FAIL("Cannot load test dataset lab3.csv!");
 
-  FFN<MeanSquaredError<>, RandomInitialization> model;
+  FFN<MeanSquaredError, RandomInitialization> model;
 
   model.Add<Linear>(2);
   model.Add<Sigmoid>();
@@ -99,8 +99,8 @@ TEST_CASE("RNNCallbackTest", "[CallbackTest]")
   RandomInitialization init(0.5, 0.5);
 
   // Create model with user defined rho parameter.
-  RNN<NegativeLogLikelihood<>, RandomInitialization> model(
-      rho, false, NegativeLogLikelihood<>(), init);
+  RNN<NegativeLogLikelihood, RandomInitialization> model(
+      rho, false, NegativeLogLikelihood(), init);
   model.Add<Linear>(10);
 
   // Use LSTM layer with 3 units.
@@ -124,8 +124,8 @@ TEST_CASE("RNNWithOptimizerCallbackTest", "[CallbackTest]")
   RandomInitialization init(0.5, 0.5);
 
   // Create model with user defined rho parameter.
-  RNN<NegativeLogLikelihood<>, RandomInitialization> model(
-      rho, false, NegativeLogLikelihood<>(), init);
+  RNN<NegativeLogLikelihood, RandomInitialization> model(
+      rho, false, NegativeLogLikelihood(), init);
   model.Add<Linear>(10);
 
   // Use LSTM layer with 3 units.

src/mlpack/tests/convolutional_network_test.cpp

@@ -81,7 +81,7 @@ TEST_CASE("PaddingTest", "[ConvolutionalNetworktest]")
   X.load("mnist_first250_training_4s_and_9s.arm");
 
   // Create the network.
-  FFN<NegativeLogLikelihood<>, RandomInitialization> model;
+  FFN<NegativeLogLikelihood, RandomInitialization> model;
 
   model.Add<Padding>(1, 2, 3, 4);
 
@@ -129,7 +129,7 @@ TEST_CASE("MaxPoolingTest", "[ConvolutionalNetworkTest]")
   X.col(2) = arma::vec("3, 4, 1, -1, 5, 5, 5, 5");
 
   // Create the network.
-  FFN<NegativeLogLikelihood<>, RandomInitialization> model;
+  FFN<NegativeLogLikelihood, RandomInitialization> model;
   model.Add<MaxPooling>(2, 2);
 
   arma::mat results;
@@ -204,7 +204,7 @@ TEST_CASE("VanillaNetworkTest", "[ConvolutionalNetworkTest]")
   bool success = false;
   for (size_t trial = 0; trial < 5; ++trial)
   {
-    FFN<NegativeLogLikelihood<>, RandomInitialization> model;
+    FFN<NegativeLogLikelihood, RandomInitialization> model;
 
     model.Add<Convolution>(8, 5, 5, 1, 1, 0, 0);
     model.Add<ReLU>();
@@ -253,7 +253,7 @@ TEST_CASE("VanillaNetworkTest", "[ConvolutionalNetworkTest]")
 
 TEST_CASE("VanillaNetworkBatchSizeTest", "[ConvolutionalNetworkTest]")
 {
-  FFN<NegativeLogLikelihood<>, RandomInitialization> model;
+  FFN<NegativeLogLikelihood, RandomInitialization> model;
 
   model.Add<Convolution>(8, 5, 5, 1, 1, 0, 0);
   model.Add<ReLU>();
@@ -400,8 +400,8 @@ TEST_CASE("CheckCopyVanillaNetworkTest", "[ConvolutionalNetworkTest]")
   // of iterations using random weights. If this works 1 of 5 times, I'm fine
   // with that. All I want to know is that the network is able to escape from
   // local minima and to solve the task.
-  FFN<NegativeLogLikelihood<>, RandomInitialization> *model =
-      new FFN<NegativeLogLikelihood<>, RandomInitialization>;
+  FFN<NegativeLogLikelihood, RandomInitialization> *model =
+      new FFN<NegativeLogLikelihood, RandomInitialization>;
 
   model->Add<Convolution>(8, 5, 5, 1, 1, 0, 0);
   model->Add<ReLU>();
@@ -417,8 +417,8 @@ TEST_CASE("CheckCopyVanillaNetworkTest", "[ConvolutionalNetworkTest]")
   model->Add<LogSoftMax>();
   model->InputDimensions() = std::vector<size_t>({ 28, 28 });
 
-  FFN<NegativeLogLikelihood<>, RandomInitialization> *model1 =
-      new FFN<NegativeLogLikelihood<>, RandomInitialization>;
+  FFN<NegativeLogLikelihood, RandomInitialization> *model1 =
+      new FFN<NegativeLogLikelihood, RandomInitialization>;
 
   model1->Add<Convolution>(8, 5, 5, 1, 1, 0, 0);
   model1->Add<ReLU>();

src/mlpack/tests/cv_test.cpp

@@ -220,7 +220,7 @@ TEST_CASE("MSEMatResponsesTest", "[CVTest]")
   arma::mat data("1 2");
   arma::mat trainingResponses("1 2; 3 4");
 
-  FFN<MeanSquaredError<>, ConstInitialization> ffn(MeanSquaredError<>(),
+  FFN<MeanSquaredError, ConstInitialization> ffn(MeanSquaredError(),
     ConstInitialization(0));
   ffn.Add<Linear>(2);
 

src/mlpack/tests/feedforward_network_2_test.cpp

@@ -102,7 +102,7 @@ TEST_CASE("RBFNetworkTest", "[FeedForwardNetworkTest]")
   KMeans<> kmeans;
   kmeans.Cluster(trainData, 8, centroids);
 
-  FFN<MeanSquaredError<> > model;
+  FFN<MeanSquaredError> model;
   model.Add<RBF>(8, centroids);
   model.Add<Linear>(3);
 
@@ -134,7 +134,7 @@ TEST_CASE("RBFNetworkTest", "[FeedForwardNetworkTest]")
   KMeans<> kmeans1;
   kmeans1.Cluster(dataset, 140, centroids1);
 
-  FFN<MeanSquaredError<> > model1;
+  FFN<MeanSquaredError> model1;
   model1.Add<RBF>(140, centroids1, 4.1);
   model1.Add<Linear>(2);
 

src/mlpack/tests/feedforward_network_test.cpp

@@ -137,13 +137,13 @@ TEST_CASE("CheckCopyMovingVanillaNetworkTest", "[FeedForwardNetworkTest]")
    * +-----+       +-----+
    */
 
-  FFN<NegativeLogLikelihood<> > *model = new FFN<NegativeLogLikelihood<> >;
+  FFN<NegativeLogLikelihood> *model = new FFN<NegativeLogLikelihood>;
   model->Add<Linear>(8);
   model->Add<Sigmoid>();
   model->Add<Linear>(3);
   model->Add<LogSoftMax>();
 
-  FFN<NegativeLogLikelihood<> > *model1 = new FFN<NegativeLogLikelihood<> >;
+  FFN<NegativeLogLikelihood> *model1 = new FFN<NegativeLogLikelihood>;
   model1->Add<Linear>(8);
   model1->Add<Sigmoid>();
   model1->Add<Linear>(3);
@@ -171,12 +171,12 @@ TEST_CASE("CheckCopyMovingReparametrizationNetworkTest",
 
   // Construct a feed forward network with trainData.n_rows input nodes,
   // followed by a linear layer and then a reparametrization layer.
-  FFN<NegativeLogLikelihood<> > *model = new FFN<NegativeLogLikelihood<> >;
+  FFN<NegativeLogLikelihood> *model = new FFN<NegativeLogLikelihood>;
   model->Add<Linear>(8);
   model->Add<Reparametrization>(false, true, 1);
   model->Add<LogSoftMax>();
 
-  FFN<NegativeLogLikelihood<> > *model1 = new FFN<NegativeLogLikelihood<> >;
+  FFN<NegativeLogLikelihood> *model1 = new FFN<NegativeLogLikelihood>;
   model1->Add<Linear>(8);
   model1->Add<Reparametrization>(false, true, 1);
   model1->Add<LogSoftMax>();
@@ -208,12 +208,12 @@ TEST_CASE("CheckCopyMovingReparametrizationNetworkTest",
 //    * followed by a linear layer and then a reparametrization layer.
 //    */
 //
-//   FFN<NegativeLogLikelihood<> > *model = new FFN<NegativeLogLikelihood<> >;
+//   FFN<NegativeLogLikelihood> *model = new FFN<NegativeLogLikelihood>;
 //   model->Add<Linear<> >(trainData.n_rows, 8);
 //   model->Add<Reparametrization<> >(4, false, true, 1);
 //   model->Add<LogSoftMax<> >();
 //
-//   FFN<NegativeLogLikelihood<> > *model1 = new FFN<NegativeLogLikelihood<> >;
+//   FFN<NegativeLogLikelihood> *model1 = new FFN<NegativeLogLikelihood>;
 //   model1->Add<Linear<> >(trainData.n_rows, 8);
 //   model1->Add<Reparametrization<> >(4, false, true, 1);
 //   model1->Add<LogSoftMax<> >();
@@ -240,13 +240,13 @@ TEST_CASE("CheckCopyMovingLinear3DNetworkTest", "[FeedForwardNetworkTest]")
 
   // Construct a feed forward network with trainData.n_rows input nodes,
   // followed by a linear layer and then a Linear3D layer.
-  FFN<NegativeLogLikelihood<> > *model = new FFN<NegativeLogLikelihood<> >;
+  FFN<NegativeLogLikelihood> *model = new FFN<NegativeLogLikelihood>;
   model->Add<Linear>(8);
   model->Add<Sigmoid>();
   model->Add<Linear3D>(3);
   model->Add<LogSoftMax>();
 
-  FFN<NegativeLogLikelihood<> > *model1 = new FFN<NegativeLogLikelihood<> >;
+  FFN<NegativeLogLikelihood> *model1 = new FFN<NegativeLogLikelihood>;
   model1->Add<Linear>(8);
   model1->Add<Sigmoid>();
   model1->Add<Linear3D>(3);
@@ -270,7 +270,7 @@ TEST_CASE("CheckCopyMovingNoisyLinearTest", "[FeedForwardNetworkTest]")
   arma::mat output = arma::mat("0");
 
   // Check copying constructor.
-  FFN<NegativeLogLikelihood<>> *model1 = new FFN<NegativeLogLikelihood<>>();
+  FFN<NegativeLogLikelihood> *model1 = new FFN<NegativeLogLikelihood>();
   model1->ResetData(input, output);
   model1->Add<NoisyLinear>(5);
   model1->Add<Linear>(1);
@@ -280,7 +280,7 @@ TEST_CASE("CheckCopyMovingNoisyLinearTest", "[FeedForwardNetworkTest]")
   CheckCopyFunction(model1, input, output);
 
   // Check moving constructor.
-  FFN<NegativeLogLikelihood<>> *model2 = new FFN<NegativeLogLikelihood<>>();
+  FFN<NegativeLogLikelihood> *model2 = new FFN<NegativeLogLikelihood>();
   model2->ResetData(input, output);
   model2->Add<NoisyLinear>(5);
   model2->Add<Linear>(1);
@@ -301,7 +301,7 @@ TEST_CASE("CheckCopyMovingConcatenateTest", "[FeedForwardNetworkTest]")
   arma::mat output = arma::mat("1");
 
   // Check copying constructor.
-  FFN<NegativeLogLikelihood<>> *model1 = new FFN<NegativeLogLikelihood<>>();
+  FFN<NegativeLogLikelihood> *model1 = new FFN<NegativeLogLikelihood>();
   model1->ResetData(input, output);
   model1->Add<Linear>(5);
 
@@ -319,7 +319,7 @@ TEST_CASE("CheckCopyMovingConcatenateTest", "[FeedForwardNetworkTest]")
   CheckCopyFunction(model1, input, output);
 
   // Check moving constructor.
-  FFN<NegativeLogLikelihood<>> *model2 = new FFN<NegativeLogLikelihood<>>();
+  FFN<NegativeLogLikelihood> *model2 = new FFN<NegativeLogLikelihood>();
   model2->ResetData(input, output);
   model2->Add<Linear>(5);
 
@@ -349,14 +349,14 @@ TEST_CASE("CheckCopyMovingDropoutNetworkTest", "[FeedForwardNetworkTest]")
   arma::mat trainLabels = trainData.row(trainData.n_rows - 1) - 1;
   trainData.shed_row(trainData.n_rows - 1);
 
-  FFN<NegativeLogLikelihood<> > *model = new FFN<NegativeLogLikelihood<> >;
+  FFN<NegativeLogLikelihood> *model = new FFN<NegativeLogLikelihood>;
   model->Add<Linear>(8);
   model->Add<Sigmoid>();
   model->Add<Dropout>(0.3);
   model->Add<Linear>(3);
   model->Add<LogSoftMax>();
 
-  FFN<NegativeLogLikelihood<> > *model1 = new FFN<NegativeLogLikelihood<> >;
+  FFN<NegativeLogLikelihood> *model1 = new FFN<NegativeLogLikelihood>;
   model1->Add<Linear>(8);
   model1->Add<Sigmoid>();
   model1->Add<Dropout>(0.3);
@@ -398,13 +398,13 @@ TEST_CASE("CheckCopyMovingVanillaNetworkTestNoBias", "[FeedForwardNetworkTest]")
    * +-----+       +--+--+       +-----+
    */
 
-  FFN<NegativeLogLikelihood<> > *model = new FFN<NegativeLogLikelihood<> >;
+  FFN<NegativeLogLikelihood> *model = new FFN<NegativeLogLikelihood>;
   model->Add<LinearNoBias>(8);
   model->Add<Sigmoid>();
   model->Add<LinearNoBias>(3);
   model->Add<LogSoftMax>();
 
-  FFN<NegativeLogLikelihood<> > *model1 = new FFN<NegativeLogLikelihood<> >;
+  FFN<NegativeLogLikelihood> *model1 = new FFN<NegativeLogLikelihood>;
   model1->Add<LinearNoBias>(8);
   model1->Add<Sigmoid>();
   model1->Add<LinearNoBias>(3);
@@ -436,12 +436,12 @@ TEST_CASE("CheckCopyMovingVanillaNetworkTestNoBias", "[FeedForwardNetworkTest]")
 //    * followed by a linear layer and then a reparametrization layer.
 //    */
 //
-//   FFN<NegativeLogLikelihood<> > *model = new FFN<NegativeLogLikelihood<> >;
+//   FFN<NegativeLogLikelihood> *model = new FFN<NegativeLogLikelihood>;
 //   model->Add<LinearNoBias<> >(trainData.n_rows, 8);
 //   model->Add<Reparametrization<> >(4, false, true, 1);
 //   model->Add<LogSoftMax<> >();
 //
-//   FFN<NegativeLogLikelihood<> > *model1 = new FFN<NegativeLogLikelihood<> >;
+//   FFN<NegativeLogLikelihood> *model1 = new FFN<NegativeLogLikelihood>;
 //   model1->Add<LinearNoBias<> >(trainData.n_rows, 8);
 //   model1->Add<Reparametrization<> >(4, false, true, 1);
 //   model1->Add<LogSoftMax<> >();
@@ -497,7 +497,7 @@ TEST_CASE("FFVanillaNetworkTest", "[FeedForwardNetworkTest]")
    * +-----+       +-----+
    */
 
-  FFN<NegativeLogLikelihood<> > model;
+  FFN<NegativeLogLikelihood> model;
   model.Add<Linear>(8);
   model.Add<Sigmoid>();
   model.Add<Linear>(3);
@@ -518,7 +518,7 @@ TEST_CASE("FFVanillaNetworkTest", "[FeedForwardNetworkTest]")
   arma::mat labels = arma::zeros(1, dataset.n_cols);
   labels.submat(0, labels.n_cols / 2, 0, labels.n_cols - 1).fill(1);
 
-  FFN<NegativeLogLikelihood<> > model1;
+  FFN<NegativeLogLikelihood> model1;
   model1.Add<Linear>(10);
   model1.Add<Sigmoid>();
   model1.Add<Linear>(2);
@@ -539,7 +539,7 @@ TEST_CASE("ForwardBackwardTest", "[FeedForwardNetworkTest]")
   arma::mat labels = arma::zeros(1, dataset.n_cols);
   labels.submat(0, labels.n_cols / 2, 0, labels.n_cols - 1).fill(1);
 
-  FFN<NegativeLogLikelihood<> > model;
+  FFN<NegativeLogLikelihood> model;
   model.Add<Linear>(50);
   model.Add<Sigmoid>();
   model.Add<Linear>(10);
@@ -645,7 +645,7 @@ TEST_CASE("DropoutNetworkTest", "[FeedForwardNetworkTest]")
    * +-----+
    */
 
-  FFN<NegativeLogLikelihood<> > model;
+  FFN<NegativeLogLikelihood> model;
   model.Add<Linear>(8);
   model.Add<Sigmoid>();
   model.Add<Dropout>();
@@ -668,7 +668,7 @@ TEST_CASE("DropoutNetworkTest", "[FeedForwardNetworkTest]")
   arma::mat labels = arma::zeros(1, dataset.n_cols);
   labels.submat(0, labels.n_cols / 2, 0, labels.n_cols - 1).fill(1);
 
-  FFN<NegativeLogLikelihood<> > model1;
+  FFN<NegativeLogLikelihood> model1;
   model1.Add<Linear>(10);
   model1.Add<Sigmoid>();
   model.Add<Dropout>();
@@ -693,7 +693,7 @@ TEST_CASE("HighwayNetworkTest", "[FeedForwardNetworkTest]")
   arma::mat labels = arma::zeros(1, dataset.n_cols);
   labels.submat(0, labels.n_cols / 2, 0, labels.n_cols - 1).fill(1);
 
-  FFN<NegativeLogLikelihood<> > model;
+  FFN<NegativeLogLikelihood> model;
   model.Add<Linear>(10);
   Highway* highway = new Highway();
   highway->Add<Linear>(10);
@@ -750,7 +750,7 @@ TEST_CASE("DropConnectNetworkTest", "[FeedForwardNetworkTest]")
   *
   */
 
-  FFN<NegativeLogLikelihood<> > model;
+  FFN<NegativeLogLikelihood> model;
   model.Add<Linear>(8);
   model.Add<Sigmoid>();
   model.Add<DropConnect>(3);
@@ -771,7 +771,7 @@ TEST_CASE("DropConnectNetworkTest", "[FeedForwardNetworkTest]")
   arma::mat labels = arma::zeros(1, dataset.n_cols);
   labels.submat(0, labels.n_cols / 2, 0, labels.n_cols - 1).fill(1);
 
-  FFN<NegativeLogLikelihood<> > model1;
+  FFN<NegativeLogLikelihood> model1;
   model1.Add<Linear>(10);
   model1.Add<Sigmoid>();
   model1.Add<DropConnect>(2);
@@ -787,7 +787,7 @@ TEST_CASE("DropConnectNetworkTest", "[FeedForwardNetworkTest]")
  */
 TEST_CASE("FFNMiscTest", "[FeedForwardNetworkTest]")
 {
-  FFN<MeanSquaredError<>> model;
+  FFN<MeanSquaredError> model;
   model.Add<Linear>(3);
   model.Add<ReLU>();
 
@@ -822,7 +822,7 @@ TEST_CASE("FFSerializationTest", "[FeedForwardNetworkTest]")
   // Vanilla neural net with logistic activation function.
   // Because 92% of the patients are not hyperthyroid the neural
   // network must be significant better than 92%.
-  FFN<NegativeLogLikelihood<> > model;
+  FFN<NegativeLogLikelihood> model;
   model.Add<Linear>(8);
   model.Add<Sigmoid>();
   model.Add<Dropout>();
@@ -833,7 +833,7 @@ TEST_CASE("FFSerializationTest", "[FeedForwardNetworkTest]")
 
   model.Train(trainData, trainLabels, opt);
 
-  FFN<NegativeLogLikelihood<>> xmlModel, jsonModel, binaryModel;
+  FFN<NegativeLogLikelihood> xmlModel, jsonModel, binaryModel;
   xmlModel.Add<Linear>(10); // Layer that will get removed.
 
   // Serialize into other models.
@@ -874,7 +874,7 @@ TEST_CASE("FFSerializationTest", "[FeedForwardNetworkTest]")
 //   // Vanilla neural net with logistic activation function.
 //   // Because 92% of the patients are not hyperthyroid the neural
 //   // network must be significant better than 92%.
-//   FFN<NegativeLogLikelihood<> > model;
+//   FFN<NegativeLogLikelihood> model;
 //   model.Add<Linear<> >(trainData.n_rows, 8);
 //   model.Add<PReLU<> >();
 //   model.Add<Dropout<> >();
@@ -885,7 +885,7 @@ TEST_CASE("FFSerializationTest", "[FeedForwardNetworkTest]")
 //
 //   model.Train(trainData, trainLabels, opt);
 //
-//   FFN<NegativeLogLikelihood<>> xmlModel, jsonModel, binaryModel;
+//   FFN<NegativeLogLikelihood> xmlModel, jsonModel, binaryModel;
 //   xmlModel.Add<Linear<>>(10, 10); // Layer that will get removed.
 //
 //   // Serialize into other models.
@@ -924,7 +924,7 @@ TEST_CASE("FFSerializationTest", "[FeedForwardNetworkTest]")
 //   testData.shed_row(testData.n_rows - 1);
 //   testLabels -= 1; // The labels should be between 0 and numClasses - 1.
 //
-//   FFN<NegativeLogLikelihood<>, RandomInitialization, CustomLayer<> > model;
+//   FFN<NegativeLogLikelihood, RandomInitialization, CustomLayer<> > model;
 //   model.Add<Linear<> >(trainData.n_rows, 8);
 //   model.Add<CustomLayer<> >();
 //   model.Add<Linear<> >(8, 3);
@@ -943,7 +943,7 @@ TEST_CASE("FFSerializationTest", "[FeedForwardNetworkTest]")
  */
 TEST_CASE("PartialForwardTest", "[FeedForwardNetworkTest]")
 {
-  FFN<NegativeLogLikelihood<>, RandomInitialization> model;
+  FFN<NegativeLogLikelihood, RandomInitialization> model;
   model.Add<Linear>(10);
 
   // Add a new Add<> module which adds a (learnable) constant term to the input.
@@ -1012,7 +1012,7 @@ TEST_CASE("FFNTrainReturnObjective", "[FeedForwardNetworkTest]")
   // Vanilla neural net with logistic activation function.
   // Because 92% of the patients are not hyperthyroid the neural
   // network must be significantly better than 92%.
-  FFN<NegativeLogLikelihood<> > model;
+  FFN<NegativeLogLikelihood> model;
   model.Add<Linear>(8);
   model.Add<Sigmoid>();
   model.Add<Dropout>();
@@ -1032,7 +1032,7 @@ TEST_CASE("FFNTrainReturnObjective", "[FeedForwardNetworkTest]")
 TEST_CASE("FFNReturnModel", "[FeedForwardNetworkTest]")
 {
   // Create dummy network.
-  FFN<NegativeLogLikelihood<> > model;
+  FFN<NegativeLogLikelihood> model;
   Linear* linearA = new Linear(3);
   model.Add(linearA);
   Linear* linearB = new Linear(4);
@@ -1082,7 +1082,7 @@ TEST_CASE("OptimizerTest", "[FeedForwardNetworkTest]")
   testData.shed_row(testData.n_rows - 1);
   testLabels -= 1; // The labels should be between 0 and numClasses.
 
-  FFN<NegativeLogLikelihood<>, RandomInitialization> model;
+  FFN<NegativeLogLikelihood, RandomInitialization> model;
   model.Add<Linear>(8);
   model.Add<Linear>(3);
   model.Add<LogSoftMax>();
@@ -1112,7 +1112,7 @@ TEST_CASE("FFNCheckInputShapeTest", "[FeedForwardNetworkTest]")
   arma::mat testLabels = testData.row(testData.n_rows - 1);
   testData.shed_row(testData.n_rows - 1);
 
-  FFN<NegativeLogLikelihood<>, RandomInitialization> model;
+  FFN<NegativeLogLikelihood, RandomInitialization> model;
   model.Add<Linear>(8);
   model.Add<Linear>(3);
   model.Add<LogSoftMax>();

src/mlpack/tests/init_rules_test.cpp

@@ -214,13 +214,13 @@ TEST_CASE("NetworkInitTest", "[InitRulesTest]")
 {
   arma::mat input = arma::ones(5, 1);
   arma::mat response;
-  NegativeLogLikelihood<> outputLayer;
+  NegativeLogLikelihood outputLayer;
 
   // Create a simple network and use the RandomInitialization rule to
   // initialize the network parameters.
   RandomInitialization randomInit(0.5, 0.5);
 
-  FFN<NegativeLogLikelihood<>, RandomInitialization> randomModel(
+  FFN<NegativeLogLikelihood, RandomInitialization> randomModel(
       std::move(outputLayer), randomInit);
   randomModel.Add<Linear>(5);
   randomModel.Add<Linear>(2);
@@ -233,7 +233,7 @@ TEST_CASE("NetworkInitTest", "[InitRulesTest]")
 
   // Create a simple network and use the OrthogonalInitialization rule to
   // initialize the network parameters.
-  FFN<NegativeLogLikelihood<>, OrthogonalInitialization> orthogonalModel;
+  FFN<NegativeLogLikelihood, OrthogonalInitialization> orthogonalModel;
   orthogonalModel.Add<Linear>(5);
   orthogonalModel.Add<Linear>(2);
   orthogonalModel.Add<LogSoftMax>();
@@ -243,8 +243,8 @@ TEST_CASE("NetworkInitTest", "[InitRulesTest]")
 
   // Create a simple network and use the ZeroInitialization rule to
   // initialize the network parameters.
-  FFN<NegativeLogLikelihood<>, ConstInitialization>
-    zeroModel(NegativeLogLikelihood<>(), ConstInitialization(0));
+  FFN<NegativeLogLikelihood, ConstInitialization>
+    zeroModel(NegativeLogLikelihood(), ConstInitialization(0));
   zeroModel.Add<Linear>(5);
   zeroModel.Add<Linear>(2);
   zeroModel.Add<LogSoftMax>();
@@ -258,7 +258,7 @@ TEST_CASE("NetworkInitTest", "[InitRulesTest]")
   // parameters.
   KathirvalavakumarSubavathiInitialization kathirvalavakumarSubavathiInit(
       input, 1.5);
-  FFN<NegativeLogLikelihood<>, KathirvalavakumarSubavathiInitialization>
+  FFN<NegativeLogLikelihood, KathirvalavakumarSubavathiInitialization>
       ksModel(std::move(outputLayer), kathirvalavakumarSubavathiInit);
   ksModel.Add<Linear>(5);
   ksModel.Add<Linear>(2);
@@ -269,7 +269,7 @@ TEST_CASE("NetworkInitTest", "[InitRulesTest]")
 
   // Create a simple network and use the OivsInitialization rule to
   // initialize the network parameters.
-  FFN<NegativeLogLikelihood<>, OivsInitialization<> > oivsModel;
+  FFN<NegativeLogLikelihood, OivsInitialization<> > oivsModel;
   oivsModel.Add<Linear>(5);
   oivsModel.Add<Linear>(2);
   oivsModel.Add<LogSoftMax>();
@@ -279,7 +279,7 @@ TEST_CASE("NetworkInitTest", "[InitRulesTest]")
 
   // Create a simple network and use the GaussianInitialization rule to
   // initialize the network parameters.
-  FFN<NegativeLogLikelihood<>, GaussianInitialization> gaussianModel;
+  FFN<NegativeLogLikelihood, GaussianInitialization> gaussianModel;
   gaussianModel.Add<Linear>(5);
   gaussianModel.Add<Linear>(2);
   gaussianModel.Add<LogSoftMax>();

src/mlpack/tests/ksinit_test.cpp

@@ -75,8 +75,8 @@ void BuildVanillaNetwork(MatType& trainData,
   // Cauchy’s Inequality Based on Sensitivity Analysis" paper.
   KathirvalavakumarSubavathiInitialization init(trainData, 4.59);
 
-  FFN<MeanSquaredError<>, KathirvalavakumarSubavathiInitialization>
-      model(MeanSquaredError<>(), init);
+  FFN<MeanSquaredError, KathirvalavakumarSubavathiInitialization>
+      model(MeanSquaredError(), init);
 
   model.Add<Linear>(hiddenLayerSize);
   model.Add<LeakyReLU>();

src/mlpack/tests/loss_functions_test.cpp

@@ -53,7 +53,7 @@ using namespace mlpack::ann;
 TEST_CASE("HuberLossTest", "[LossFunctionsTest]")
 {
   arma::mat input, target, output;
-  HuberLoss<> module;
+  HuberLoss module;
 
   // Test the Forward function.
   input = arma::mat("17.45 12.91 13.63 29.01 7.12 15.47 31.52 31.97");
@@ -82,10 +82,10 @@ TEST_CASE("PoissonNLLLossTest", "[LossFunctionsTest]")
   arma::mat input, target, input4, target4;
   arma::mat output1, output2, output3, output4;
   arma::mat expOutput1, expOutput2, expOutput3, expOutput4;
-  PoissonNLLLoss<> module1;
-  PoissonNLLLoss<> module2(true, true, 1e-08, false);
-  PoissonNLLLoss<> module3(true, true, 1e-08, true);
-  PoissonNLLLoss<> module4(false, true, 1e-08, true);
+  PoissonNLLLoss module1;
+  PoissonNLLLoss module2(true, true, 1e-08, false);
+  PoissonNLLLoss module3(true, true, 1e-08, true);
+  PoissonNLLLoss module4(false, true, 1e-08, true);
 
   // Test the Forward function on a user generated input.
   input = arma::mat("1.0 1.0 1.9 1.6 -1.9 3.7 -1.0 0.5");
@@ -148,7 +148,7 @@ TEST_CASE("SimpleKLDivergenceTest", "[LossFunctionsTest]")
 {
   arma::mat input, target, output;
   double loss;
-  KLDivergence<> module(true);
+  KLDivergence module(true);
 
   // Test the Forward function.  Loss should be 0 if input = target.
   input = arma::ones(10, 1);
@@ -163,7 +163,7 @@ TEST_CASE("SimpleKLDivergenceTest", "[LossFunctionsTest]")
 TEST_CASE("SimpleMeanSquaredLogarithmicErrorTest", "[LossFunctionsTest]")
 {
   arma::mat input, output, target;
-  MeanSquaredLogarithmicError<> module;
+  MeanSquaredLogarithmicError module;
 
   // Test the Forward function on a user generator input and compare it against
   // the manually calculated result.
@@ -198,7 +198,7 @@ TEST_CASE("KLDivergenceMeanTest", "[LossFunctionsTest]")
 {
   arma::mat input, target, output;
   double loss;
-  KLDivergence<> module(true);
+  KLDivergence module(true);
 
   // Test the Forward function.
   input = arma::mat("1 1 1 1 1 1 1 1 1 1");
@@ -219,7 +219,7 @@ TEST_CASE("KLDivergenceNoMeanTest", "[LossFunctionsTest]")
 {
   arma::mat input, target, output;
   double loss;
-  KLDivergence<> module(false);
+  KLDivergence module(false);
 
   // Test the Forward function.
   input = arma::mat("1 1 1 1 1 1 1 1 1 1");
@@ -239,7 +239,7 @@ TEST_CASE("KLDivergenceNoMeanTest", "[LossFunctionsTest]")
 TEST_CASE("SimpleMeanSquaredErrorTest", "[LossFunctionsTest]")
 {
   arma::mat input, output, target;
-  MeanSquaredError<> module;
+  MeanSquaredError module;
 
   // Test the Forward function on a user generator input and compare it against
   // the manually calculated result.
@@ -276,8 +276,8 @@ TEST_CASE("SimpleMeanSquaredErrorTest", "[LossFunctionsTest]")
 TEST_CASE("SimpleBinaryCrossEntropyLossTest", "[LossFunctionsTest]")
 {
   arma::mat input1, input2, input3, output, target1, target2, target3;
-  BCELoss<> module1(1e-6, false);
-  BCELoss<> module2(1e-6, true);
+  BCELoss module1(1e-6, false);
+  BCELoss module2(1e-6, true);
   // Test the Forward function on a user generator input and compare it against
   // the manually calculated result.
   input1 = arma::mat("0.5 0.5 0.5 0.5 0.5 0.5 0.5 0.5");
@@ -329,7 +329,7 @@ TEST_CASE("SimpleSigmoidCrossEntropyErrorTest", "[LossFunctionsTest]")
 {
   arma::mat input1, input2, input3, output, target1,
             target2, target3, expectedOutput;
-  SigmoidCrossEntropyError<> module;
+  SigmoidCrossEntropyError module;
 
   // Test the Forward function on a user generator input and compare it against
   // the calculated result.
@@ -388,7 +388,7 @@ TEST_CASE("SimpleSigmoidCrossEntropyErrorTest", "[LossFunctionsTest]")
 TEST_CASE("SimpleEarthMoverDistanceLayerTest", "[LossFunctionsTest]")
 {
   arma::mat input1, input2, output, target1, target2, expectedOutput;
-  EarthMoverDistance<> module;
+  EarthMoverDistance module;
 
   // Test the Forward function on a user generator input and compare it against
   // the manually calculated result.
@@ -433,7 +433,7 @@ TEST_CASE("GradientMeanSquaredErrorTest", "[LossFunctionsTest]")
       input = arma::randu(10, 1);
       target = arma::randu(2, 1);
 
-      model = new FFN<MeanSquaredError<>, NguyenWidrowInitialization>();
+      model = new FFN<MeanSquaredError, NguyenWidrowInitialization>();
       model->ResetData(input, target);
       model->Add<Linear>(2);
       model->Add<Sigmoid>();
@@ -454,7 +454,7 @@ TEST_CASE("GradientMeanSquaredErrorTest", "[LossFunctionsTest]")
 
     arma::mat& Parameters() { return model->Parameters(); }
 
-    FFN<MeanSquaredError<>, NguyenWidrowInitialization>* model;
+    FFN<MeanSquaredError, NguyenWidrowInitialization>* model;
     arma::mat input, target;
   } function;
 
@@ -474,7 +474,7 @@ TEST_CASE("GradientReconstructionLossTest", "[LossFunctionsTest]")
       input = arma::randu(10, 1);
       target = arma::randu(2, 1);
 
-      model = new FFN<ReconstructionLoss<>, NguyenWidrowInitialization>();
+      model = new FFN<ReconstructionLoss, NguyenWidrowInitialization>();
       model->ResetData(input, target);
       model->Add<Linear>(2);
       model->Add<Sigmoid>();
@@ -495,7 +495,7 @@ TEST_CASE("GradientReconstructionLossTest", "[LossFunctionsTest]")
 
     arma::mat& Parameters() { return model->Parameters(); }
 
-    FFN<ReconstructionLoss<>, NguyenWidrowInitialization>* model;
+    FFN<ReconstructionLoss, NguyenWidrowInitialization>* model;
     arma::mat input, target;
   } function;
 
@@ -509,7 +509,7 @@ TEST_CASE("DiceLossTest", "[LossFunctionsTest]")
 {
   arma::mat input1, input2, target, output;
   double loss;
-  DiceLoss<> module;
+  DiceLoss module;
 
   // Test the Forward function. Loss should be 0 if input = target.
   input1 = arma::ones(10, 1);
@@ -549,7 +549,7 @@ TEST_CASE("DiceLossTest", "[LossFunctionsTest]")
 TEST_CASE("SimpleMeanBiasErrorTest", "[LossFunctionsTest]")
 {
   arma::mat input, output, target;
-  MeanBiasError<> module;
+  MeanBiasError module;
 
   // Test the Forward function on a user generator input and compare it against
   // the manually calculated result.
@@ -588,7 +588,7 @@ TEST_CASE("LogCoshLossTest", "[LossFunctionsTest]")
 {
   arma::mat input, target, output;
   double loss;
-  LogCoshLoss<> module(2);
+  LogCoshLoss module(2);
 
   // Test the Forward function. Loss should be 0 if input = target.
   input = arma::ones(10, 1);
@@ -627,7 +627,7 @@ TEST_CASE("HingeEmbeddingLossTest", "[LossFunctionsTest]")
 {
   arma::mat input, target, output;
   double loss;
-  HingeEmbeddingLoss<> module;
+  HingeEmbeddingLoss module;
 
   // Test the Forward function. Loss should be 0 if input = target.
   input = arma::ones(10, 1);
@@ -665,7 +665,7 @@ TEST_CASE("HingeEmbeddingLossTest", "[LossFunctionsTest]")
 TEST_CASE("SimpleL1LossTest", "[LossFunctionsTest]")
 {
   arma::mat input1, input2, output, target1, target2;
-  L1Loss<> module(false);
+  L1Loss module(false);
 
   // Test the Forward function on a user generator input and compare it against
   // the manually calculated result.
@@ -702,7 +702,7 @@ TEST_CASE("CosineEmbeddingLossTest", "[LossFunctionsTest]")
 {
   arma::mat input1, input2, y, output;
   double loss;
-  CosineEmbeddingLoss<> module;
+  CosineEmbeddingLoss module;
 
   // Test the Forward function. Loss should be 0 if input1 = input2 and y = 1.
   input1 = arma::mat(1, 10);
@@ -750,7 +750,7 @@ TEST_CASE("CosineEmbeddingLossTest", "[LossFunctionsTest]")
 TEST_CASE("MarginRankingLossTest", "[LossFunctionsTest]")
 {
   arma::mat input, input1, input2, target, output;
-  MarginRankingLoss<> module;
+  MarginRankingLoss module;
 
   // Test the Forward function on a user generator input and compare it against
   // the manually calculated result.
@@ -794,8 +794,8 @@ TEST_CASE("SoftMarginLossTest", "[LossFunctionsTest]")
 {
   arma::mat input, target, output, expectedOutput;
   double loss;
-  SoftMarginLoss<> module1;
-  SoftMarginLoss<> module2(false);
+  SoftMarginLoss module1;
+  SoftMarginLoss module2(false);
 
   input = arma::mat("0.1778 0.0957 0.1397 0.1203 0.2403 0.1925 -0.2264 -0.3400 "
       "-0.3336");
@@ -849,7 +849,7 @@ TEST_CASE("SoftMarginLossTest", "[LossFunctionsTest]")
 TEST_CASE("MeanAbsolutePercentageErrorTest", "[LossFunctionsTest]")
 {
   arma::mat input, target, output, expectedOutput;
-  MeanAbsolutePercentageError<> module;
+  MeanAbsolutePercentageError module;
 
   input = arma::mat("3 -0.5 2 7");
   target = arma::mat("2.5 0.2 2 8");
@@ -876,7 +876,7 @@ TEST_CASE("MeanAbsolutePercentageErrorTest", "[LossFunctionsTest]")
 TEST_CASE("VRClassRewardLayerParametersTest", "[LossFunctionsTest]")
 {
   // Parameter order : scale, sizeAverage.
-  VRClassReward<> layer(2, false);
+  VRClassReward layer(2, false);
 
   // Make sure we can get the parameters successfully.
   REQUIRE(layer.Scale() == 2);
@@ -890,7 +890,7 @@ TEST_CASE("TripletMarginLossTest")
 {
   arma::mat anchor, positive, negative;
   arma::mat input, target, output;
-  TripletMarginLoss<> module;
+  TripletMarginLoss module;
 
   // Test the Forward function on a user generated input and compare it against
   // the manually calculated result.
@@ -938,8 +938,8 @@ TEST_CASE("HingeLossTest", "[LossFunctionsTest]")
 {
   arma::mat input, target, target_b, output;
   double loss, loss_b;
-  HingeLoss<> module1;
-  HingeLoss<> module2(false);
+  HingeLoss module1;
+  HingeLoss module2(false);
 
   // Test the Forward function. Loss should be 0 if input = target.
   input = arma::ones(10, 1);
@@ -1015,8 +1015,8 @@ TEST_CASE("MultiLabelSoftMarginLossTest", "[LossFunctionsTest]")
 {
   arma::mat input, target, output, expectedOutput;
   double loss;
-  MultiLabelSoftMarginLoss<> module1;
-  MultiLabelSoftMarginLoss<> module2(false);
+  MultiLabelSoftMarginLoss module1;
+  MultiLabelSoftMarginLoss module2(false);
 
   input = arma::mat("0.1778 0.0957 0.1397 0.1203 0.2403 0.1925 -0.2264 -0.3400 "
       "-0.3336");
@@ -1074,8 +1074,8 @@ TEST_CASE("MultiLabelSoftMarginLossWeightedTest", "[LossFunctionsTest]")
   arma::rowvec weights;
   double loss;
   weights = arma::mat("1 2 3");
-  MultiLabelSoftMarginLoss<> module1(true, weights);
-  MultiLabelSoftMarginLoss<> module2(false, weights);
+  MultiLabelSoftMarginLoss module1(true, weights);
+  MultiLabelSoftMarginLoss module2(false, weights);
 
   input = arma::mat("0.1778 0.0957 0.1397 0.2256 0.1203 0.2403 0.1925 0.3144 "
       "-0.2264 -0.3400 -0.3336 -0.8695");

src/mlpack/tests/recurrent_network_test.cpp

@@ -571,7 +571,7 @@ void GenerateNoisySinRNN(arma::cube& data,
  */
 double RNNSineTest(size_t hiddenUnits, size_t rho, size_t numEpochs = 100)
 {
-  RNN<MeanSquaredError<> > net(rho, true);
+  RNN<MeanSquaredError> net(rho, true);
   net.Add<LinearNoBias>(hiddenUnits);
   net.Add<LSTM>(hiddenUnits);
   net.Add<LinearNoBias>(1);
@@ -853,8 +853,8 @@ TEST_CASE("RNNFFNTest", "[RecurrentNetworkTest]")
 {
   // We'll create an RNN with *no* BPTT, just a simple single-layer linear
   // network.
-  RNN<MeanSquaredError<>, ConstInitialization> rnn;
-  FFN<MeanSquaredError<>, ConstInitialization> ffn;
+  RNN<MeanSquaredError, ConstInitialization> rnn;
+  FFN<MeanSquaredError, ConstInitialization> ffn;
 
   rnn.Add<Linear>(10);
   rnn.Add<Sigmoid>();