Update RBF<> layer so tests pass.

src/mlpack/methods/ann/layer/layer_types.hpp

@@ -58,7 +58,7 @@
 //#include <mlpack/methods/ann/layer/padding.hpp>
 //#include <mlpack/methods/ann/layer/parametric_relu.hpp>
 //#include <mlpack/methods/ann/layer/positional_encoding.hpp>
-//#include <mlpack/methods/ann/layer/radial_basis_function.hpp>
+#include <mlpack/methods/ann/layer/radial_basis_function.hpp>
 //#include <mlpack/methods/ann/layer/recurrent.hpp>
 //#include <mlpack/methods/ann/layer/recurrent_attention.hpp>
 //#include <mlpack/methods/ann/layer/reinforce_normal.hpp>

src/mlpack/methods/ann/layer/radial_basis_function.hpp

@@ -68,6 +68,9 @@ class RBF : public Layer<InputType, OutputType>
       InputType& centres,
       double betas = 0);
 
+  //! Clone the LinearType object. This handles polymorphism correctly.
+  RBF* Clone() const { return new RBF(*this); }
+
   /**
    * Ordinary feed forward pass of the radial basis function.
    *
@@ -83,12 +86,12 @@ class RBF : public Layer<InputType, OutputType>
                 const OutputType& /* gy */,
                 OutputType& /* g */);
 
-  const std::vector<size_t>& OutputDimensions() const
+  void ComputeOutputDimensions()
   {
-    std::vector<size_t> outputDimensions(inputDimensions.size(), 1);
+    this->outputDimensions = std::vector<size_t>(this->inputDimensions.size(),
+        1);
     // This flattens the input.
-    outputDimensions[0] = outSize;
-    return outputDimensions;
+    this->outputDimensions[0] = outSize;
   }
 
   /**

src/mlpack/methods/ann/layer/radial_basis_function_impl.hpp

@@ -2,7 +2,6 @@
  * @file radial_basis_function_impl.hpp
  * @author Himanshu Pathak
  *
- *
  * 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
  * 3-clause BSD license along with mlpack.  If not, see
@@ -96,7 +95,7 @@ void RBF<InputType, OutputType, Activation>::serialize(
 
   ar(CEREAL_NVP(distances));
   ar(CEREAL_NVP(centres));
-  ar(CEREAL_NVP(betas);
+  ar(CEREAL_NVP(betas));
 }
 
 } // namespace ann

src/mlpack/methods/ann/layer/serialization.hpp

@@ -125,6 +125,7 @@
     CEREAL_REGISTER_TYPE(mlpack::ann::NoisyLinearType<__VA_ARGS__>); \
     CEREAL_REGISTER_TYPE(mlpack::ann::ConcatenateType<__VA_ARGS__>); \
     CEREAL_REGISTER_TYPE(mlpack::ann::AddType<__VA_ARGS__>); \
+    CEREAL_REGISTER_TYPE(mlpack::ann::RBF<__VA_ARGS__>); \
 
 // TODO: continue...
 

src/mlpack/tests/CMakeLists.txt

@@ -35,6 +35,7 @@ add_executable(mlpack_test
   # facilities_test.cpp
   # fastmks_test.cpp
   feedforward_network_test.cpp
+  feedforward_network_2_test.cpp
   # gan_test.cpp
   # gmm_test.cpp
   # hmm_test.cpp

src/mlpack/tests/custom_layer.hpp

@@ -15,7 +15,8 @@
 
 #include <mlpack/prereqs.hpp>
 #include <mlpack/methods/ann/layer/layer.hpp>
-
+#include <mlpack/methods/ann/layer/base_layer.hpp>
+#include <mlpack/methods/ann/activation_functions/logistic_function.hpp>
 
 namespace mlpack {
 namespace ann {

src/mlpack/tests/feedforward_network_2_test.cpp

@@ -14,6 +14,7 @@
 #include <mlpack/core.hpp>
 
 #include <mlpack/methods/ann/layer/layer.hpp>
+#include <mlpack/methods/ann/layer/layer_types.hpp>
 #include <mlpack/methods/ann/loss_functions/mean_squared_error.hpp>
 #include <mlpack/methods/ann/ffn.hpp>
 #include <mlpack/methods/kmeans/kmeans.hpp>
@@ -100,8 +101,8 @@ TEST_CASE("RBFNetworkTest", "[FeedForwardNetworkTest]")
   kmeans.Cluster(trainData, 8, centroids);
 
   FFN<MeanSquaredError<> > model;
-  model.Add<RBF<> >(trainData.n_rows, 8, centroids);
-  model.Add<Linear<> >(8, 3);
+  model.Add<RBF<> >(8, centroids);
+  model.Add<Linear>(3);
 
   // RBFN neural net with MeanSquaredError.
   TestNetwork<>(model, trainData, trainLabels1, testData, testLabels, 10, 0.1);
@@ -132,8 +133,8 @@ TEST_CASE("RBFNetworkTest", "[FeedForwardNetworkTest]")
   kmeans1.Cluster(dataset, 140, centroids1);
 
   FFN<MeanSquaredError<> > model1;
-  model1.Add<RBF<> >(dataset.n_rows, 140, centroids1, 4.1);
-  model1.Add<Linear<> >(140, 2);
+  model1.Add<RBF<> >(140, centroids1, 4.1);
+  model1.Add<Linear>(2);
 
   // RBFN neural net with MeanSquaredError.
   TestNetwork<>(model1, dataset, labels1, dataset, labels, 10, 0.1);