adding locally_connected_1d to neural-fortran environment

Author: Riccardo231

example/cnn_mnist_1d.f90

@@ -1,7 +1,7 @@
 program cnn_mnist
 
     use nf, only: network, sgd, &
-      input, conv2d, maxpool2d, flatten, dense, reshape, &
+      input, conv2d, maxpool2d, flatten, dense, reshape, locally_connected_1d, &
       load_mnist, label_digits, softmax, relu
 
     implicit none
@@ -20,11 +20,8 @@ program cnn_mnist
 
     net = network([ &
       input(784), &
-      reshape([1,28,28]), &
-      conv2d(filters=8, kernel_size=3, activation=relu()), &
-      maxpool2d(pool_size=2), &
-      conv2d(filters=16, kernel_size=3, activation=relu()), &
-      maxpool2d(pool_size=2), &
+      reshape([1,784]), &
+      locally_connected_1d(filters=8, kernel_size=2, activation=relu()), &
       dense(10, activation=softmax()) &
     ])
 

src/nf.f90

@@ -3,7 +3,7 @@ module nf
   use nf_datasets_mnist, only: label_digits, load_mnist
   use nf_layer, only: layer
   use nf_layer_constructors, only: &
-    conv2d, dense, flatten, input, maxpool2d, reshape
+    conv2d, dense, flatten, input, maxpool2d, reshape, locally_connected_1d
   use nf_loss, only: mse, quadratic
   use nf_metrics, only: corr, maxabs
   use nf_network, only: network

src/nf/nf_layer_constructors.f90

@@ -8,7 +8,7 @@ module nf_layer_constructors
   implicit none
 
   private
-  public :: conv2d, dense, flatten, input, maxpool2d, reshape
+  public :: conv2d, dense, flatten, input, locally_connected_1d, maxpool2d, reshape
 
   interface input
 
@@ -174,6 +174,33 @@ module function maxpool2d(pool_size, stride) result(res)
         !! Resulting layer instance
     end function maxpool2d
 
+    module function locally_connected_1d(filters, kernel_size, activation) result(res)
+      !! 2-d convolutional layer constructor.
+      !!
+      !! This layer is for building 2-d convolutional network.
+      !! Although the established convention is to call these layers 2-d,
+      !! the shape of the data is actuall 3-d: image width, image height,
+      !! and the number of channels.
+      !! A conv2d layer must not be the first layer in the network.
+      !!
+      !! Example:
+      !!
+      !! ```
+      !! use nf, only :: conv2d, layer
+      !! type(layer) :: conv2d_layer
+      !! conv2d_layer = dense(filters=32, kernel_size=3)
+      !! conv2d_layer = dense(filters=32, kernel_size=3, activation='relu')
+      !! ```
+      integer, intent(in) :: filters
+        !! Number of filters in the output of the layer
+      integer, intent(in) :: kernel_size
+        !! Width of the convolution window, commonly 3 or 5
+      class(activation_function), intent(in), optional :: activation
+        !! Activation function (default sigmoid)
+      type(layer) :: res
+        !! Resulting layer instance
+    end function locally_connected_1d
+
     module function reshape(output_shape) result(res)
       !! Rank-1 to rank-any reshape layer constructor.
       !! Currently implemented is only rank-3 for the output of the reshape.

src/nf/nf_layer_constructors_submodule.f90

@@ -7,6 +7,7 @@
   use nf_input1d_layer, only: input1d_layer
   use nf_input2d_layer, only: input2d_layer
   use nf_input3d_layer, only: input3d_layer
+  use nf_locally_connected_1d_layer, only: locally_connected_1d_layer
   use nf_maxpool2d_layer, only: maxpool2d_layer
   use nf_reshape_layer, only: reshape3d_layer
   use nf_activation, only: activation_function, relu, sigmoid
@@ -103,6 +104,33 @@ module function input3d(dim1, dim2, dim3) result(res)
     res % initialized = .true.
   end function input3d
 
+<<<<<<< HEAD
+=======
+  module function locally_connected_1d(filters, kernel_size, activation) result(res)
+    integer, intent(in) :: filters
+    integer, intent(in) :: kernel_size
+    class(activation_function), intent(in), optional :: activation
+    type(layer) :: res
+    class(activation_function), allocatable :: activation_tmp
+
+    res % name = 'locally_connected_1d'
+
+    if (present(activation)) then
+      allocate(activation_tmp, source=activation)
+    else
+      allocate(activation_tmp, source=relu())
+    end if
+
+    res % activation = activation_tmp % get_name()
+
+    allocate( &
+      res % p, &
+      source=locally_connected_1d_layer(filters, kernel_size, activation_tmp) &
+    )
+
+  end function locally_connected_1d
+    
+>>>>>>> 0fbbda4 (adding locally_connected_1d to neural-fortran environment)
 
   module function maxpool2d(pool_size, stride) result(res)
     integer, intent(in) :: pool_size

src/nf/nf_network_submodule.f90

@@ -6,10 +6,11 @@
   use nf_input1d_layer, only: input1d_layer
   use nf_input2d_layer, only: input2d_layer
   use nf_input3d_layer, only: input3d_layer
+  use nf_locally_connected_1d_layer, only: locally_connected_1d_layer
   use nf_maxpool2d_layer, only: maxpool2d_layer
   use nf_reshape_layer, only: reshape3d_layer
   use nf_layer, only: layer
-  use nf_layer_constructors, only: conv2d, dense, flatten, input, maxpool2d, reshape
+  use nf_layer_constructors, only: conv2d, dense, flatten, input, locally_connected_1d, maxpool2d, reshape
   use nf_loss, only: quadratic
   use nf_optimizers, only: optimizer_base_type, sgd
   use nf_parallel, only: tile_indices

src/nf/nf_reshape_layer_submodule.f90

@@ -48,4 +48,6 @@ module subroutine init(self, input_shape)
 
   end subroutine init
 
+  
+
 end submodule nf_reshape_layer_submodule