fnn_network_sequential Module Reference

Module dedicated to the class sequentialneuralnetwork. More...

Data Types
type	layercontainer
	Layer container class. More...
type	sequentialneuralnetwork
	Implements a sequential neural network. More...

Functions/Subroutines
type(sequentialneuralnetwork) function, public	construct_sequential_neural_network (batch_size, internal_sizes, activation_names, initialisation_names)
	Manual constructor for class sequentialneuralnetwork. Only for testing purpose. More...
type(sequentialneuralnetwork) function, public	snn_fromfile (batch_size, filename)
	Constructor for class sequentialneuralnetwork from a file. More...
integer(ik) function	snn_get_input_size (self)
	Implements sequentialneuralnetwork::get_input_size. More...
integer(ik) function	snn_get_output_size (self)
	Implements sequentialneuralnetwork::get_output_size. More...
integer(ik) function	snn_get_num_parameters (self)
	Implements sequentialneuralnetwork::get_num_parameters. More...
subroutine	snn_set_parameters (self, new_parameters)
	Implements sequentialneuralnetwork::set_parameters. More...
subroutine	snn_get_parameters (self, parameters)
	Implements sequentialneuralnetwork::get_parameters. More...
subroutine	snn_tofile (self, filename)
	Implements sequentialneuralnetwork::tofile. More...
subroutine	snn_apply_forward (self, train, member, x, y)
	Implements sequentialneuralnetwork::apply_forward. More...
subroutine	snn_apply_tangent_linear (self, member, dp, dx, dy)
	Implements sequentialneuralnetwork::apply_tangent_linear. More...
subroutine	snn_apply_adjoint (self, member, dy, dp, dx)
	Implements sequentialneuralnetwork::apply_adjoint. More...

Detailed Description

Module dedicated to the class sequentialneuralnetwork.

Function/Subroutine Documentation

construct_sequential_neural_network()

type(sequentialneuralnetwork) function, public fnn_network_sequential::construct_sequential_neural_network	(	integer(ik), intent(in)	batch_size,
		integer(ik), dimension(:), intent(in)	internal_sizes,
		character(len=*), dimension(:), intent(in)	activation_names,
		character(len=*), dimension(:), intent(in)	initialisation_names
	)

Manual constructor for class sequentialneuralnetwork. Only for testing purpose.

Parameters

[in]	batch_size	The value for sequentialneuralnetwork::batch_size.
[in]	internal_sizes	The list of internal sizes.
[in]	activation_names	The list of activation functions.
[in]	initialisation_names	The list of initialisations for model parameters.

Returns

The constructed network.

snn_apply_adjoint()

subroutine fnn_network_sequential::snn_apply_adjoint ( class(sequentialneuralnetwork), intent(inout)  self, integer(ik), intent(in)  member, real(rk), dimension(:), intent(inout)  dy, real(rk), dimension(:), intent(out)  dp, real(rk), dimension(:), intent(out)  dx  )

private

Implements sequentialneuralnetwork::apply_adjoint.

Applies the adjoint of the network.

Note

If there is more than one layer, this method uses layer::adjoint_input as intermediate storage.

For this reason, the intent of self is declared inout. In addition, the intent of dy is declared intout because for certain layer classes (e.g. fnn_layer_dense::denselayer) the output perturbation gets overwritten in the process.

Parameters

[in,out]	self	The network.
[in]	member	The index inside the batch.
[in,out]	dy	The output perturbation.
[out]	dp	The parameter perturbation.
[out]	dx	The state perturbation.

snn_apply_forward()

subroutine fnn_network_sequential::snn_apply_forward ( class(sequentialneuralnetwork), intent(inout)  self, logical, intent(in)  train, integer(ik), intent(in)  member, real(rk), dimension(:), intent(in)  x, real(rk), dimension(:), intent(out)  y  )

private

Implements sequentialneuralnetwork::apply_forward.

Applies and linearises the network.

Note

If there is more than one layer, this method uses layer::forward_input as intermediate storage.

For the case of a fnn_layer_dense::denselayer, this means that the first step of denselayer::apply_forward consists in copying layer::forward_input into... layer::forward_input!

For theses reasons, and because the linearisation is stored inside the network, the intent of self is declared inout.

Parameters

[in,out]	self	The network.
[in]	train	Whether the model is used in training mode.
[in]	member	The index inside the batch.
[in]	x	The input of the network.
[out]	y	The output of the network.

snn_apply_tangent_linear()

subroutine fnn_network_sequential::snn_apply_tangent_linear ( class(sequentialneuralnetwork), intent(inout)  self, integer(ik), intent(in)  member, real(rk), dimension(:), intent(in)  dp, real(rk), dimension(:), intent(in)  dx, real(rk), dimension(:), intent(out)  dy  )

private

Implements sequentialneuralnetwork::apply_tangent_linear.

Applies the TL of the network.

Note

If there is more than one layer, this method uses layer::tangent_linear_input as intermediate storage.

For this reason, the intent of self is declared inout.

Parameters

[in,out]	self	The network.
[in]	member	The index inside the batch.
[in]	dp	The parameter perturbation.
[in]	dx	The state perturbation.
[out]	dy	The output perturbation.

snn_fromfile()

type(sequentialneuralnetwork) function, public fnn_network_sequential::snn_fromfile	(	integer(ik), intent(in)	batch_size,
		character(len=*), intent(in)	filename
	)

Constructor for class sequentialneuralnetwork from a file.

Parameters

[in]	batch_size	The value for layer::batch_size.
[in]	filename	The name of the file to read.

Returns

The constructed network.

snn_get_input_size()

integer(ik) function fnn_network_sequential::snn_get_input_size ( class(sequentialneuralnetwork), intent(in)  self )

private

Implements sequentialneuralnetwork::get_input_size.

Returns the input size of the network.

Parameters

[in]

self

The network.

Returns

The input size.

snn_get_num_parameters()

integer(ik) function fnn_network_sequential::snn_get_num_parameters ( class(sequentialneuralnetwork), intent(in)  self )

private

Implements sequentialneuralnetwork::get_num_parameters.

Returns the number of parameters.

Parameters

[in]

self

The network.

Returns

The number of parameters.

snn_get_output_size()

integer(ik) function fnn_network_sequential::snn_get_output_size ( class(sequentialneuralnetwork), intent(in)  self )

private

Implements sequentialneuralnetwork::get_output_size.

Returns the output size of the network.

Parameters

[in]

self

The network.

Returns

The output size.

snn_get_parameters()

subroutine fnn_network_sequential::snn_get_parameters ( class(sequentialneuralnetwork), intent(in)  self, real(rk), dimension(:), intent(out)  parameters  )

private

Implements sequentialneuralnetwork::get_parameters.

Getter for the network's parameters.

Parameters

[in]	self	The network.
[out]	parameters	The vector of parameters.

snn_set_parameters()

subroutine fnn_network_sequential::snn_set_parameters ( class(sequentialneuralnetwork), intent(inout)  self, real(rk), dimension(:), intent(in)  new_parameters  )

private

Implements sequentialneuralnetwork::set_parameters.

Setter for the network's parameters.

Parameters

[in,out]	self	The network.
[in]	new_parameters	The new values for the parameters.

snn_tofile()

subroutine fnn_network_sequential::snn_tofile ( class(sequentialneuralnetwork), intent(in)  self, character(len=*), intent(in)  filename  )

private

Implements sequentialneuralnetwork::tofile.

Saves the network.

Parameters

[in]	self	The network.
[in]	filename	The name of the file to write.