modified as suggested by Milan

Author: Vandenplas, Jeremie

src/lib/mod_network.f90

@@ -26,23 +26,22 @@ module mod_network
 
     procedure, public, pass(self) :: accuracy
     procedure, public, pass(self) :: backprop
-    procedure, public, pass(self) :: fit_batch
     procedure, public, pass(self) :: fwdprop
     procedure, public, pass(self) :: init
     procedure, public, pass(self) :: load
     procedure, public, pass(self) :: loss
-    procedure, public, pass(self) :: output
-    procedure, public, pass(self) :: predict_batch
+    procedure, public, pass(self) :: output_batch
+    procedure, public, pass(self) :: output_single
     procedure, public, pass(self) :: save
     procedure, public, pass(self) :: set_activation
     procedure, public, pass(self) :: sync
     procedure, public, pass(self) :: train_batch
+    procedure, public, pass(self) :: train_epochs
     procedure, public, pass(self) :: train_single
     procedure, public, pass(self) :: update
 
-    generic, public :: fit => fit_batch!, train_single
-    generic, public :: predict => predict_batch!, train_single
-    generic, public :: train => train_batch, train_single
+    generic, public :: output => output_batch, output_single
+    generic, public :: train => train_batch, train_epochs, train_single
 
   end type network_type
 
@@ -121,44 +120,6 @@ pure subroutine backprop(self, y, dw, db)
 
   end subroutine backprop
 
-  subroutine fit_batch(self, x, y, eta,epochs,batch_size)
-    !Performs the training for n epochs with mini-bachtes of size equal to batch_size
-    class(network_type), intent(in out) :: self
-    integer(ik),intent(in),optional::epochs,batch_size
-    real(rk), intent(in) :: x(:,:), y(:,:), eta
-
-    integer(ik)::i,n,nsamples,nbatch
-    integer(ik)::num_epochs,num_batch_size
-    integer(ik)::batch_start,batch_end
-
-    real(rk)::pos
-
-    nsamples=size(y,dim=2)
-
-    num_epochs=1
-    if(present(epochs))num_epochs=epochs
-
-    num_batch_size=nsamples
-    if(present(batch_size))num_batch_size=batch_size
-
-    nbatch=nsamples/num_batch_size
-
-    epoch: do n=1,num_epochs
-     mini_batches: do i=1,nbatch
-      
-      !pull a random mini-batch from the dataset  
-      call random_number(pos)
-      batch_start=int(pos*(nsamples-num_batch_size+1))
-      if(batch_start.eq.0)batch_start=1
-      batch_end=batch_start+batch_size-1
-   
-      call self%train(x(:,batch_start:batch_end),y(:,batch_start:batch_end),eta)
-       
-     enddo mini_batches
-    enddo epoch
-
-  end subroutine
-
   pure subroutine fwdprop(self, x)
     ! Performs the forward propagation and stores arguments to activation
     ! functions and activations themselves for use in backprop.
@@ -216,7 +177,7 @@ pure real(rk) function loss(self, x, y)
     loss = 0.5 * sum((y - self % output(x))**2) / size(x)
   end function loss
 
-  pure function output(self, x) result(a)
+  pure function output_single(self, x) result(a)
     ! Use forward propagation to compute the output of the network.
     class(network_type), intent(in) :: self
     real(rk), intent(in) :: x(:)
@@ -228,9 +189,9 @@ pure function output(self, x) result(a)
         a = self % activation(matmul(transpose(layers(n-1) % w), a) + layers(n) % b)
       end do
     end associate
-  end function output
+  end function output_single
 
-  pure function predict_batch(self, x) result(a)
+  pure function output_batch(self, x) result(a)
     class(network_type), intent(in) :: self
     real(rk), intent(in) :: x(:,:)
     real(rk), allocatable :: a(:,:)
@@ -242,7 +203,7 @@ pure function predict_batch(self, x) result(a)
      a(:,i)=self%output(x(:,i))
     enddo
 
-  end function predict_batch
+  end function output_batch
 
   subroutine save(self, filename)
     ! Saves the network to a file.
@@ -342,6 +303,37 @@ subroutine train_batch(self, x, y, eta)
 
   end subroutine train_batch
 
+  subroutine train_epochs(self, x, y, eta,num_epochs,num_batch_size)
+    !Performs the training for nun_epochs epochs with mini-bachtes of size equal to num_batch_size
+    class(network_type), intent(in out) :: self
+    integer(ik),intent(in)::num_epochs,num_batch_size
+    real(rk), intent(in) :: x(:,:), y(:,:), eta
+
+    integer(ik)::i,n,nsamples,nbatch
+    integer(ik)::batch_start,batch_end
+
+    real(rk)::pos
+
+    nsamples=size(y,dim=2)
+
+    nbatch=nsamples/num_batch_size
+
+    epoch: do n=1,num_epochs
+     mini_batches: do i=1,nbatch
+      
+      !pull a random mini-batch from the dataset  
+      call random_number(pos)
+      batch_start=int(pos*(nsamples-num_batch_size+1))
+      if(batch_start.eq.0)batch_start=1
+      batch_end=batch_start+num_batch_size-1
+   
+      call self%train(x(:,batch_start:batch_end),y(:,batch_start:batch_end),eta)
+       
+     enddo mini_batches
+    enddo epoch
+
+  end subroutine train_epochs
+
   pure subroutine train_single(self, x, y, eta)
     ! Trains a network using a single set of input data x and output data y,
     ! and learning rate eta.

src/tests/example_mnist.f90

@@ -30,7 +30,7 @@ program example_mnist
       net % accuracy(te_images, label_digits(te_labels)) * 100, ' %'
   end if
 
-  call net%fit(tr_images,label_digits(tr_labels),eta=3._rk,epochs=num_epochs,batch_size=batch_size)
+  call net%train(tr_images,label_digits(tr_labels),3._rk,num_epochs,batch_size)
 
   if (this_image() == 1) then
     write(*, '(a,f5.2,a)') 'Epochs done, Accuracy: ',&

src/tests/example_montesinos_multi.f90

@@ -21,7 +21,7 @@ program example_montesinos_multi
  num_epochs=50
 
  !training
- call net%fit(x_tr,y_tr,3._rk,epochs=num_epochs,batch_size=batch_size)
+ call net%train(x_tr,y_tr,3._rk,num_epochs,batch_size)
 
  call net%sync(1)
 
@@ -30,7 +30,7 @@ program example_montesinos_multi
  call readfile('../data/montesinos_multi/x_ts.dat',nx1_ts,nx2_ts,x_ts)
 
  if(this_image().eq.1)then
-  write(*,*)'Correlation(s): ',corr_array(net%predict(x_ts),y_ts)
+  write(*,*)'Correlation(s): ',corr_array(net%output(x_ts),y_ts)
  endif
 
 contains

src/tests/example_montesinos_uni.f90

@@ -22,7 +22,7 @@ program example_montesinos_uni
  num_epochs=20
 
  !training
- call net%fit(x_tr,y_tr,3._rk,epochs=num_epochs,batch_size=batch_size)
+ call net%train(x_tr,y_tr,3._rk,num_epochs,batch_size)
 
  call net%sync(1)
 
@@ -31,7 +31,7 @@ program example_montesinos_uni
  call readfile('../data/montesinos_uni/x_ts.dat',nx1_ts,nx2_ts,x_ts)
 
  if(this_image().eq.1)then
-  write(*,*)'Correlation(s): ',corr_array(net%predict(x_ts),y_ts)
+  write(*,*)'Correlation(s): ',corr_array(net%output(x_ts),y_ts)
  endif
 
 contains