MSE is Float32 not Float64

spec/network_spec.cr

@@ -18,6 +18,7 @@ describe Fann::Network do
 
   it "trains on single data" do
     ann = Fann::Network::Standard.new(2, [2, 2], 1)
+    ann.randomzie_weights(0.0, 1.0)
     3000.times do
       ann.train_single([1.0, 0.0], [0.5])
     end
@@ -33,10 +34,11 @@ describe Fann::Network do
 
   it "trains and evaluate single data" do
     ann = Fann::Network::Standard.new(2, [2], 1)
+    ann.randomzie_weights(0.0, 1.0)
     3000.times do
-      ann.train_single([1.0, 0.1], [0.5])
+      ann.train_single([1.0, 0.0], [0.5])
     end
-    result = ann.run([1.0, 0.1])
+    result = ann.run([1.0, 0.0])
     ann.close
     (result < [0.55] && result > [0.45]).should be_true
   end
@@ -47,9 +49,10 @@ describe Fann::Network do
     output = [[0.0], [1.0], [1.0], [0.0]]
     train_data = Fann::TrainData.new(input, output)
     data = train_data.train_data
-    ann.train_algorithem(LibFANN::TrainEnum::TrainRprop)
-    ann.set_hidden_layer_activation_func(LibFANN::ActivationfuncEnum::LeakyRelu)
-    ann.set_output_layer_activation_func(LibFANN::ActivationfuncEnum::LeakyRelu)
+    # ann.train_algorithem(LibFANN::TrainEnum::TrainSarprop)
+    ann.randomzie_weights(0.0, 1.0)
+    # ann.set_hidden_layer_activation_func(LibFANN::ActivationfuncEnum::LeakyRelu)
+    # ann.set_output_layer_activation_func(LibFANN::ActivationfuncEnum::LeakyRelu)
     if data
       ann.train_batch(data, {:max_runs => 8000, :desired_mse => 0.001, :log_each => 1000})
     end
@@ -65,8 +68,9 @@ describe Fann::Network do
     train_data = Fann::TrainData.new(input, output)
     data = train_data.train_data
     ann.train_algorithem(LibFANN::TrainEnum::TrainRprop)
+    ann.randomzie_weights(0.0, 1.0)
     if data
-      ann.train_batch(data, {:max_neurons => 500, :desired_mse => 0.1, :log_each => 10})
+      ann.train_batch(data, {:max_neurons => 500, :desired_mse => 0.001, :log_each => 10})
     end
     result = ann.run([1.0, 1.0])
     ann.close

src/crystal-fann/cascade_network.cr

@@ -29,14 +29,14 @@ module Fann
         LibFANN.set_activation_function_hidden(@nn, func)
       end
 
-      def randomzie_weights(min : Float32, max : Float32)
+      def randomzie_weights(min : Float64, max : Float64)
         # randomize_weights = fann_randomize_weights(ann : Fann*, min_weight : Type, max_weight : Type)
         LibFANN.randomize_weights(@nn, min, max)
       end
 
       def train_batch(train_data : Pointer(LibFANN::TrainData), opts = {:max_neurons => 500, :desired_mse => 0.01_f64, :log_each => 10})
         # fun cascadetrain_on_data = fann_cascadetrain_on_data(ann : Fann*, data : TrainData*, max_neurons : LibC::UInt, neurons_between_reports : LibC::UInt, desired_error : LibC::Double)
-        LibFANN.cascade_train_on_data(@nn, train_data, opts[:max_neurons], opts[:log_each], opts[:desired_mse])
+        LibFANN.cascade_train_on_data(@nn, train_data, opts[:max_neurons], opts[:log_each], opts[:desired_mse].to_f32)
       end
 
       def run(input : Array(Float64))

src/crystal-fann/lib_fann.cr

@@ -292,8 +292,8 @@ lib LibFANN
   fun get_mse = fann_get_MSE(ann : Fann*) : LibC::Double
   fun get_bit_fail = fann_get_bit_fail(ann : Fann*) : LibC::UInt
   fun reset_mse = fann_reset_MSE(ann : Fann*)
-  fun train_on_data = fann_train_on_data(ann : Fann*, data : TrainData*, max_epochs : LibC::UInt, epochs_between_reports : LibC::UInt, desired_error : LibC::Double)
-  fun train_on_file = fann_train_on_file(ann : Fann*, filename : LibC::Char*, max_epochs : LibC::UInt, epochs_between_reports : LibC::UInt, desired_error : LibC::Double)
+  fun train_on_data = fann_train_on_data(ann : Fann*, data : TrainData*, max_epochs : LibC::UInt, epochs_between_reports : LibC::UInt, desired_error : LibC::Float)
+  fun train_on_file = fann_train_on_file(ann : Fann*, filename : LibC::Char*, max_epochs : LibC::UInt, epochs_between_reports : LibC::UInt, desired_error : LibC::Float)
   fun train_epoch = fann_train_epoch(ann : Fann*, data : TrainData*) : LibC::Double
   fun train_epoch_lw = fann_train_epoch_lw(ann : Fann*, data : TrainData*, label_weight : Type*) : LibC::Double
   fun train_epoch_irpropm_gradient = fann_train_epoch_irpropm_gradient(ann : Fann*, data : TrainData*, error_function : (Type*, Type*, LibC::Int, Void* -> Type), x3 : Void*) : LibC::Double
@@ -381,8 +381,8 @@ lib LibFANN
   fun set_sarprop_step_error_shift = fann_set_sarprop_step_error_shift(ann : Fann*, sarprop_step_error_shift : LibC::Double)
   fun get_sarprop_temperature = fann_get_sarprop_temperature(ann : Fann*) : LibC::Double
   fun set_sarprop_temperature = fann_set_sarprop_temperature(ann : Fann*, sarprop_temperature : LibC::Double)
-  fun cascade_train_on_data = fann_cascadetrain_on_data(ann : Fann*, data : TrainData*, max_neurons : LibC::UInt, neurons_between_reports : LibC::UInt, desired_error : LibC::Double)
-  fun cascade_train_on_file = fann_cascadetrain_on_file(ann : Fann*, filename : LibC::Char*, max_neurons : LibC::UInt, neurons_between_reports : LibC::UInt, desired_error : LibC::Double)
+  fun cascade_train_on_data = fann_cascadetrain_on_data(ann : Fann*, data : TrainData*, max_neurons : LibC::UInt, neurons_between_reports : LibC::UInt, desired_error : LibC::Float)
+  fun cascade_train_on_file = fann_cascadetrain_on_file(ann : Fann*, filename : LibC::Char*, max_neurons : LibC::UInt, neurons_between_reports : LibC::UInt, desired_error : LibC::Float)
   fun get_cascade_output_change_fraction = fann_get_cascade_output_change_fraction(ann : Fann*) : LibC::Double
   fun set_cascade_output_change_fraction = fann_set_cascade_output_change_fraction(ann : Fann*, cascade_output_change_fraction : LibC::Double)
   fun get_cascade_output_stagnation_epochs = fann_get_cascade_output_stagnation_epochs(ann : Fann*) : LibC::UInt

src/crystal-fann/standard_network.cr

@@ -37,7 +37,7 @@ module Fann
         LibFANN.set_activation_function_output(@nn, func)
       end
 
-      def randomzie_weights(min : Float32, max : Float32)
+      def randomzie_weights(min : Float64, max : Float64)
         # randomize_weights = fann_randomize_weights(ann : Fann*, min_weight : Type, max_weight : Type)
         LibFANN.randomize_weights(@nn, min, max)
       end
@@ -50,7 +50,7 @@ module Fann
 
       def train_batch(train_data : Pointer(LibFANN::TrainData), opts = {:max_runs => 200, :desired_mse => 0.01_f64, :log_each => 1})
         # train_on_data = fann_train_on_data(ann : Fann*, data : TrainData*, max_epochs : LibC::UInt, epochs_between_reports : LibC::UInt, desired_error : LibC::Double)
-        LibFANN.train_on_data(@nn, train_data, opts[:max_runs], opts[:log_each], opts[:desired_mse])
+        LibFANN.train_on_data(@nn, train_data, opts[:max_runs], opts[:log_each], opts[:desired_mse].to_f32)
       end
 
       def train_batch_multicore(train_data : Pointer(LibFANN::TrainData), threads : Int32, opts = {:max_runs => 200, :desired_mse => 0.01_f64, :log_each => 1})