diff --git a/src/nf/nf_initializers.f90 b/src/nf/nf_initializers.f90
new file mode 100644
index 0000000..1cf0da9
--- /dev/null
+++ b/src/nf/nf_initializers.f90
@@ -0,0 +1,44 @@
+module nf_initializers
+  implicit none
+
+  private
+  public :: initializer_type, glorot, he
+
+  type, abstract :: initializer_type
+  contains
+    procedure(init), deferred :: init
+  end type initializer_type
+
+  abstract interface
+    subroutine init(self, x)
+      import :: initializer_type
+      class(initializer_type), intent(in) :: self
+      real, intent(inout) :: x(:)
+    end subroutine init
+  end interface
+
+  type, extends(initializer_type) :: glorot
+  contains
+    procedure :: init => init_glorot
+  end type glorot
+
+  type, extends(initializer_type) :: he
+  contains
+    procedure :: init => init_he
+  end type he
+
+contains
+
+  subroutine init_glorot(self, x)
+    class(glorot), intent(in) :: self
+    real, intent(inout) :: x(:)
+    error stop 'Not implemented'
+  end subroutine init_glorot
+
+  subroutine init_he(self, x)
+    class(he), intent(in) :: self
+    real, intent(inout) :: x(:)
+    error stop 'Not implemented'
+  end subroutine init_he
+
+end module nf_initializers
\ No newline at end of file
diff --git a/src/nf/nf_layer.f90 b/src/nf/nf_layer.f90
index e9e90da..cfae0e8 100644
--- a/src/nf/nf_layer.f90
+++ b/src/nf/nf_layer.f90
@@ -4,6 +4,7 @@ module nf_layer
   !! user-facing API.
 
   use nf_base_layer, only: base_layer
+  use nf_initializers, only: initializer_type
   use nf_optimizers, only: optimizer_base_type
 
   implicit none
@@ -17,6 +18,7 @@ module nf_layer
     !! nf_layer_constructors.f90 to create `layer` instances.
 
     class(base_layer), allocatable :: p
+    class(initializer_type), allocatable :: initializer
     character(:), allocatable :: name
     character(:), allocatable :: activation
     integer, allocatable :: layer_shape(:)
diff --git a/src/nf/nf_layer_constructors.f90 b/src/nf/nf_layer_constructors.f90
index ce9a724..61c805c 100644
--- a/src/nf/nf_layer_constructors.f90
+++ b/src/nf/nf_layer_constructors.f90
@@ -3,7 +3,8 @@ module nf_layer_constructors
   !! This module provides the functions to instantiate specific layers.
 
   use nf_layer, only: layer
-  use nf_activation, only : activation_function
+  use nf_initializers, only: initializer_type
+  use nf_activation, only: activation_function
 
   implicit none
 
@@ -62,7 +63,7 @@ end function input3d
 
   interface
 
-    pure module function dense(layer_size, activation) result(res)
+    pure module function dense(layer_size, activation, initializer) result(res)
       !! Dense (fully-connected) layer constructor.
       !!
       !! This layer is a building block for dense, fully-connected networks,
@@ -81,6 +82,8 @@ pure module function dense(layer_size, activation) result(res)
         !! The number of neurons in a dense layer
       class(activation_function), intent(in), optional :: activation
         !! Activation function instance (default sigmoid)
+      class(initializer_type), intent(in), optional :: initializer
+        !! Algorithm to use to initialize the weights
       type(layer) :: res
         !! Resulting layer instance
     end function dense
@@ -106,7 +109,7 @@ pure module function flatten() result(res)
         !! Resulting layer instance
     end function flatten
 
-    pure module function conv2d(filters, kernel_size, activation) result(res)
+    pure module function conv2d(filters, kernel_size, activation, initializer) result(res)
       !! 2-d convolutional layer constructor.
       !!
       !! This layer is for building 2-d convolutional network.
@@ -121,7 +124,7 @@ pure module function conv2d(filters, kernel_size, activation) result(res)
       !! 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')
+      !! conv2d_layer = dense(filters=32, kernel_size=3, activation=relu())
       !! ```
       integer, intent(in) :: filters
         !! Number of filters in the output of the layer
@@ -129,6 +132,8 @@ pure module function conv2d(filters, kernel_size, activation) result(res)
         !! Width of the convolution window, commonly 3 or 5
       class(activation_function), intent(in), optional :: activation
         !! Activation function (default sigmoid)
+      class(initializer_type), intent(in), optional :: initializer
+        !! Algorithm to use to initialize the weights
       type(layer) :: res
         !! Resulting layer instance
     end function conv2d
diff --git a/src/nf/nf_layer_constructors_submodule.f90 b/src/nf/nf_layer_constructors_submodule.f90
index 002a83b..e9046fb 100644
--- a/src/nf/nf_layer_constructors_submodule.f90
+++ b/src/nf/nf_layer_constructors_submodule.f90
@@ -9,15 +9,17 @@
   use nf_maxpool2d_layer, only: maxpool2d_layer
   use nf_reshape_layer, only: reshape3d_layer
   use nf_activation, only: activation_function, sigmoid
+  use nf_initializers, only: initializer_type
 
   implicit none
 
 contains
 
-  pure module function conv2d(filters, kernel_size, activation) result(res)
+  pure module function conv2d(filters, kernel_size, activation, initializer) result(res)
     integer, intent(in) :: filters
     integer, intent(in) :: kernel_size
     class(activation_function), intent(in), optional :: activation
+    class(initializer_type), intent(in), optional :: initializer
     type(layer) :: res
 
     class(activation_function), allocatable :: activation_tmp
@@ -32,6 +34,8 @@ pure module function conv2d(filters, kernel_size, activation) result(res)
 
     res % activation = activation_tmp % get_name()
 
+    if (present(initializer)) error stop 'Initializers not yet implemented'
+
     allocate( &
       res % p, &
       source=conv2d_layer(filters, kernel_size, activation_tmp) &
@@ -40,9 +44,10 @@ pure module function conv2d(filters, kernel_size, activation) result(res)
   end function conv2d
 
 
-  pure module function dense(layer_size, activation) result(res)
+  pure module function dense(layer_size, activation, initializer) result(res)
     integer, intent(in) :: layer_size
     class(activation_function), intent(in), optional :: activation
+    class(initializer_type), intent(in), optional :: initializer
     type(layer) :: res
 
     class(activation_function), allocatable :: activation_tmp
@@ -58,6 +63,8 @@ pure module function dense(layer_size, activation) result(res)
 
     res % activation = activation_tmp % get_name()
 
+    if (present(initializer)) error stop 'Initializers not yet implemented'
+
     allocate(res % p, source=dense_layer(layer_size, activation_tmp))
 
   end function dense