Steepest descent optimizer with move limits. More...

Inheritance diagram for steepest_descent_optimizer::steepest_descent_optimizer_t:

Collaboration diagram for steepest_descent_optimizer::steepest_descent_optimizer_t:

Public Member Functions
procedure, pass(this)	init_from_json (this, parameters, problem, design, simulation)
	Initialise the steepest descent optimizer from the case file.

procedure, pass(this)	free (this)
	Free the steepest descent optimizer.

procedure, pass(this)	initialize (this, problem, design, simulation)
	Evaluate the problem and its sensitivity at the starting design.

procedure, pass(this)	step (this, iter, problem, design, simulation)
	Take one steepest descent step and re-evaluate the problem.

procedure, pass(this)	validate (this, problem, design)
	Validate the final design.

procedure, pass(this)	write (this, iter, problem)
	Write the progress of the optimizer to the log file.

procedure, pass(this)	save_checkpoint_components (this, filename, overwrite)
	Save the steepest descent state to a checkpoint.

procedure, pass(this)	load_checkpoint_components (this, filename)
	Restore the steepest descent state from a checkpoint.

procedure(optimizer_init_from_json), deferred, pass, public	init_from_json (this, parameters, problem, design, simulation)
	Initialize the optimizer, associate it with a specific problem.

procedure(optimizer_free), deferred, pass, public	free (this)
	Free resources.

procedure(optimizer_initialize), deferred, pass, public	initialize (this, problem, design, simulation)
	Prepare the optimizer before starting the optimization loop.

procedure(optimizer_step), deferred, pass, public	step (this, iter, problem, design, simulation)
	Perform a single optimization step.

procedure(optimizer_validate), deferred, pass, public	validate (this, problem, design)
	Validate the solution.

procedure(optimizer_write), deferred, pass, public	write (this, iter, problem)
	Write the progress of the optimizer to the log file.

procedure(optimizer_save_checkpoint_components), deferred, pass	save_checkpoint_components (this, filename, overwrite)
	Save optimizer-specific components to checkpoint.

procedure(optimizer_load_checkpoint_components), deferred, pass	load_checkpoint_components (this, filename)
	Load optimizer-specific components from checkpoint.

procedure, pass(this)	save_checkpoint (this, design, overwrite, filename, path, basename, format, iteration)
	Save a checkpoint of the optimizer state.

procedure, pass(this)	load_checkpoint (this, design, filename, path, basename, format, iteration)
	Restore the optimizer state from a checkpoint.

procedure, pass, public	run (this, problem, design, simulation)
	Run the optimization loop.

procedure, pass(this)	init_base (this, optimizer_type, max_iterations, max_runtime, stop_design_change, restart_file, checkpoint_path, checkpoint_base, checkpoint_format, checkpoint_interval, checkpoint_overwrite)
	The base initializer.

procedure, pass(this)	free_base (this)
	Free base resources.

procedure, pass(this)	read_base_settings (this, solver_params)
	Read settings from JSON parameters file.

procedure, pass(this)	get_iteration (this)
	Return the current iteration number.

procedure, pass(this)	print_status (this, stop_flag, iteration)
	Print status message.

procedure, pass(this)	out_of_time (this, step_time)
	Estimate if we are out of time.

procedure, pass(this)	init_log (this, problem, extra_headers, include_constraints, filename)
	Initialize optimization log.

procedure, pass(this)	write_log (this, iter, problem, extra_values)
	Write optimization log entry.

Public Attributes
real(kind=rp), public	max_design_change = 0.0_rp
	The maximum observed design change.

real(kind=rp), public	norm2_design_change = 0.0_rp
	The L2 norm of the design change.

Detailed Description

Definition at line 83 of file steepest_descent_optimizer.f90.

Member Function/Subroutine Documentation

◆ free() [1/2]

procedure(optimizer_free), deferred, pass, public optimizer::optimizer_t::free ( class(optimizer_t), intent(inout) this )

pure virtualinherited

Definition at line 117 of file optimizer.f90.

◆ free() [2/2]

procedure, pass(this) steepest_descent_optimizer::steepest_descent_optimizer_t::free ( class(steepest_descent_optimizer_t), intent(inout) this )

Parameters

this	The steepest descent optimizer.

Definition at line 118 of file steepest_descent_optimizer.f90.

◆ free_base()

procedure, pass(this) optimizer::optimizer_t::free_base ( class(optimizer_t), intent(inout) this )

inherited

Parameters

this	The optimizer object.

Definition at line 152 of file optimizer.f90.

◆ get_iteration()

procedure, pass(this) optimizer::optimizer_t::get_iteration ( class(optimizer_t), intent(in) this )

inherited

Parameters

this	The optimizer object.

Returns: The current iteration number.

Definition at line 156 of file optimizer.f90.

◆ init_base()

procedure, pass(this) optimizer::optimizer_t::init_base	(	class(optimizer_t), intent(inout)	this,
		character(len=*), intent(in)	optimizer_type,
		integer, intent(in)	max_iterations,
		real(kind=rp), intent(in), optional	max_runtime,
		real(kind=rp), intent(in), optional	stop_design_change,
		character(len=*), intent(in), optional	restart_file,
		character(len=*), intent(in), optional	checkpoint_path,
		character(len=*), intent(in), optional	checkpoint_base,
		character(len=*), intent(in), optional	checkpoint_format,
		integer, intent(in), optional	checkpoint_interval,
		logical, intent(in), optional	checkpoint_overwrite
	)

inherited

Interface for reading a checkpoint Base initializer for the optimizer

Parameters

this	The optimizer object.
optimizer_type	The type of the optimizer.
max_iterations	The maximum number of iterations.
max_runtime	The maximum runtime in seconds.
stop_design_change	The stopping criteria for design change.
restart_file	The checkpoint file to restart from.
checkpoint_path	The path for saving checkpoint files.
checkpoint_base	The base name for checkpoint files.
checkpoint_format	The file format for checkpoint files.
checkpoint_interval	The interval for saving checkpoints in iterations.
checkpoint_overwrite	Whether to overwrite existing checkpoint files.

Definition at line 150 of file optimizer.f90.

◆ init_from_json() [1/2]

procedure(optimizer_init_from_json), deferred, pass, public optimizer::optimizer_t::init_from_json	(	class(optimizer_t), intent(inout)	this,
		type(json_file), intent(inout)	parameters,
		class(problem_t), intent(inout)	problem,
		class(design_t), intent(in)	design,
		type(simulation_t), intent(in), optional	simulation
	)

pure virtualinherited

Definition at line 114 of file optimizer.f90.

◆ init_from_json() [2/2]

procedure, pass(this) steepest_descent_optimizer::steepest_descent_optimizer_t::init_from_json	(	class(steepest_descent_optimizer_t), intent(inout)	this,
		type(json_file), intent(inout)	parameters,
		class(problem_t), intent(inout)	problem,
		class(design_t), intent(in)	design,
		type(simulation_t), intent(in), optional	simulation
	)

Reads optimization.solver and its steepest_descent object. The base settings are read after free, which resets them, and before init_base, which keeps them.

Parameters

this	The steepest descent optimizer.
parameters	The JSON case file.
problem	The problem to optimise, which must have no constraints.
design	The design whose values start the optimisation.
simulation	The simulation, unused.

Definition at line 117 of file steepest_descent_optimizer.f90.

◆ init_log()

procedure, pass(this) optimizer::optimizer_t::init_log	(	class(optimizer_t), intent(inout)	this,
		class(problem_t), intent(in)	problem,
		character(len=*), dimension(:), intent(in), optional	extra_headers,
		logical, intent(in), optional	include_constraints,
		character(len=*), intent(in), optional	filename
	)

inherited

Parameters

[in,out]	this	The optimizer object.
[in]	problem	The problem object.
[in]	extra_headers	Header labels for extra log entries.
[in]	include_constraints	Include constraints in the log.
[in]	filename	Output filename for the log.

Note: An optimizer at iteration 0 overwrites the target log file and writes a fresh header. Otherwise, logging appends to the existing file.

Definition at line 162 of file optimizer.f90.

◆ initialize() [1/2]

procedure(optimizer_initialize), deferred, pass, public optimizer::optimizer_t::initialize	(	class(optimizer_t), intent(inout)	this,
		class(problem_t), intent(inout)	problem,
		class(design_t), intent(inout)	design,
		type(simulation_t), intent(inout), optional	simulation
	)

pure virtualinherited

Definition at line 120 of file optimizer.f90.

◆ initialize() [2/2]

procedure, pass(this) steepest_descent_optimizer::steepest_descent_optimizer_t::initialize	(	class(steepest_descent_optimizer_t), intent(inout)	this,
		class(problem_t), intent(inout)	problem,
		class(design_t), intent(inout)	design,
		type(simulation_t), intent(inout), optional	simulation
	)

Parameters

this	The steepest descent optimizer.
problem	The problem to optimise.
design	The design.
simulation	The simulation, if any.

Definition at line 119 of file steepest_descent_optimizer.f90.

◆ load_checkpoint()

procedure, pass(this) optimizer::optimizer_t::load_checkpoint	(	class(optimizer_t), intent(inout)	this,
		class(design_t), intent(inout)	design,
		character(len=*), intent(in), optional	filename,
		character(len=*), intent(in), optional	path,
		character(len=*), intent(in), optional	basename,
		character(len=*), intent(in), optional	format,
		integer, intent(in), optional	iteration
	)

inherited

Parameters

this	The optimizer object.
filename	The name of the file to load the checkpoint from.
design	The design object.
path	The path where the checkpoint file is located.
basename	The base name of the file to load the checkpoint from.
format	The file format to use for the checkpoint file.
iteration	The iteration to load.

Definition at line 138 of file optimizer.f90.

◆ load_checkpoint_components() [1/2]

procedure(optimizer_load_checkpoint_components), deferred, pass optimizer::optimizer_t::load_checkpoint_components	(	class(optimizer_t), intent(inout)	this,
		character(len=*), intent(in)	filename
	)

pure virtualinherited

Definition at line 132 of file optimizer.f90.

◆ load_checkpoint_components() [2/2]

procedure, pass(this) steepest_descent_optimizer::steepest_descent_optimizer_t::load_checkpoint_components	(	class(steepest_descent_optimizer_t), intent(inout)	this,
		character(len=*), intent(in)	filename
	)

Parameters

this	The steepest descent optimizer.
filename	The checkpoint file.

Definition at line 126 of file steepest_descent_optimizer.f90.

◆ out_of_time()

procedure, pass(this) optimizer::optimizer_t::out_of_time	(	class(optimizer_t), intent(inout)	this,
		real(kind=rp), intent(in)	step_time
	)

inherited

Parameters

this	The optimizer object.
step_time	The time taken for the latest iteration.

Returns: out_of_time Logical indicating if we are out of time.

Definition at line 160 of file optimizer.f90.

◆ print_status()

procedure, pass(this) optimizer::optimizer_t::print_status	(	class(optimizer_t), intent(in)	this,
		integer, intent(in)	stop_flag,
		integer, intent(in)	iteration
	)

inherited

Parameters

this	The optimizer object.
stop_flag	The stopping flag.
iteration	The number of iterations performed.

Definition at line 158 of file optimizer.f90.

◆ read_base_settings()

procedure, pass(this) optimizer::optimizer_t::read_base_settings	(	class(optimizer_t), intent(inout)	this,
		type(json_file), intent(inout)	solver_params
	)

inherited

Parameters

this	The optimizer object.
solver_params	The JSON file containing the optimizer parameters.

Definition at line 154 of file optimizer.f90.

◆ run()

procedure, pass, public optimizer::optimizer_t::run	(	class(optimizer_t), intent(inout)	this,
		class(problem_t), intent(inout)	problem,
		class(design_t), intent(inout)	design,
		type(simulation_t), intent(inout), optional	simulation
	)

inherited

The optimization loop can be terminated based on a maximum runtime. In this case, a cumulative average is used to determine if the next iteration would exceed the maximum runtime.

When restarting from a checkpoint, the continuation parameters are set to their values at the checkpointed iteration before the problem is re-evaluated.

Parameters

this	The optimizer object.
problem	The problem object.
design	The design object.
simulation	The simulation object.

Definition at line 144 of file optimizer.f90.

◆ save_checkpoint()

procedure, pass(this) optimizer::optimizer_t::save_checkpoint	(	class(optimizer_t), intent(inout)	this,
		class(design_t), intent(inout)	design,
		logical, intent(in), optional	overwrite,
		character(len=*), intent(in), optional	filename,
		character(len=*), intent(in), optional	path,
		character(len=*), intent(in), optional	basename,
		character(len=*), intent(in), optional	format,
		integer, intent(in), optional	iteration
	)

inherited

Parameters

this	The optimizer object.
design	The design object.
overwrite	Whether to overwrite the file if it exists.
filename	The name of the file to save the checkpoint to.
path	The path where the checkpoint file will be saved.
basename	The base name of the file to save the checkpoint.
format	The file format to use for the checkpoint file.
iteration	The iteration used to construct the checkpoint filename.

Definition at line 136 of file optimizer.f90.

◆ save_checkpoint_components() [1/2]

procedure(optimizer_save_checkpoint_components), deferred, pass optimizer::optimizer_t::save_checkpoint_components	(	class(optimizer_t), intent(inout)	this,
		character(len=*), intent(in)	filename,
		logical, intent(in), optional	overwrite
	)

pure virtualinherited

Definition at line 129 of file optimizer.f90.

◆ save_checkpoint_components() [2/2]

procedure, pass(this) steepest_descent_optimizer::steepest_descent_optimizer_t::save_checkpoint_components	(	class(steepest_descent_optimizer_t), intent(inout)	this,
		character(len=*), intent(in)	filename,
		logical, intent(in), optional	overwrite
	)

Parameters

this	The steepest descent optimizer.
filename	The checkpoint file.
overwrite	Replace an existing steepest descent group.

Definition at line 124 of file steepest_descent_optimizer.f90.

◆ step() [1/2]

procedure(optimizer_step), deferred, pass, public optimizer::optimizer_t::step	(	class(optimizer_t), intent(inout)	this,
		integer, intent(in)	iter,
		class(problem_t), intent(inout)	problem,
		class(design_t), intent(inout)	design,
		type(simulation_t), intent(inout), optional	simulation
	)

pure virtualinherited

Definition at line 122 of file optimizer.f90.

◆ step() [2/2]

procedure, pass(this) steepest_descent_optimizer::steepest_descent_optimizer_t::step	(	class(steepest_descent_optimizer_t), intent(inout)	this,
		integer, intent(in)	iter,
		class(problem_t), intent(inout)	problem,
		class(design_t), intent(inout)	design,
		type(simulation_t), intent(inout), optional	simulation
	)

The sensitivity of the current design must be up to date on entry; it is on exit, for the updated design.

Parameters

this	The steepest descent optimizer.
iter	The current iteration, starting from 1.
problem	The problem to optimise.
design	The design, updated in place.
simulation	The simulation, if any.

Returns: .true. only when the bound-projected gradient vanishes.

Definition at line 120 of file steepest_descent_optimizer.f90.

◆ validate() [1/2]

procedure(optimizer_validate), deferred, pass, public optimizer::optimizer_t::validate	(	class(optimizer_t), intent(inout)	this,
		class(problem_t), intent(in)	problem,
		class(design_t), intent(in)	design
	)

pure virtualinherited

Definition at line 124 of file optimizer.f90.

◆ validate() [2/2]

procedure, pass(this) steepest_descent_optimizer::steepest_descent_optimizer_t::validate	(	class(steepest_descent_optimizer_t), intent(inout)	this,
		class(problem_t), intent(in)	problem,
		class(design_t), intent(in)	design
	)

Nothing to check: the update keeps the design within its box bounds by construction, and problems with constraints are rejected at initialisation.

Parameters

this	The steepest descent optimizer.
problem	The problem.
design	The design.

Definition at line 121 of file steepest_descent_optimizer.f90.

◆ write() [1/2]

procedure(optimizer_write), deferred, pass, public optimizer::optimizer_t::write	(	class(optimizer_t), intent(inout)	this,
		integer, intent(in)	iter,
		class(problem_t), intent(inout)	problem
	)

pure virtualinherited

Definition at line 127 of file optimizer.f90.

◆ write() [2/2]

procedure, pass(this) steepest_descent_optimizer::steepest_descent_optimizer_t::write	(	class(steepest_descent_optimizer_t), intent(inout)	this,
		integer, intent(in)	iter,
		class(problem_t), intent(inout)	problem
	)

Parameters

this	The steepest descent optimizer.
iter	The current iteration.
problem	The problem.

Definition at line 122 of file steepest_descent_optimizer.f90.

◆ write_log()

procedure, pass(this) optimizer::optimizer_t::write_log	(	class(optimizer_t), intent(inout)	this,
		integer, intent(in)	iter,
		class(problem_t), intent(in)	problem,
		real(kind=rp), dimension(:), intent(in), optional	extra_values
	)

inherited

Parameters

[in,out]	this	The optimizer object.
[in]	iter	Current iteration number.
[in]	problem	The problem object.
[in]	extra_values	Extra log values appended after problem entries.

Definition at line 164 of file optimizer.f90.

Member Data Documentation

◆ max_design_change

real(kind=rp), public optimizer::optimizer_t::max_design_change = 0.0_rp

inherited

Definition at line 75 of file optimizer.f90.

◆ norm2_design_change

real(kind=rp), public optimizer::optimizer_t::norm2_design_change = 0.0_rp

inherited

Definition at line 77 of file optimizer.f90.

The documentation for this type was generated from the following file:

/home/runner/work/neko-top/neko-top/sources/optimizer/steepest_descent_optimizer.f90

Public Member Functions

Public Attributes

Detailed Description

Member Function/Subroutine Documentation

◆ free() [1/2]

◆ free() [2/2]

◆ free_base()

◆ get_iteration()

◆ init_base()

◆ init_from_json() [1/2]

◆ init_from_json() [2/2]

◆ init_log()

◆ initialize() [1/2]

◆ initialize() [2/2]

◆ load_checkpoint()

◆ load_checkpoint_components() [1/2]

◆ load_checkpoint_components() [2/2]

◆ out_of_time()

◆ print_status()

◆ read_base_settings()

◆ run()

◆ save_checkpoint()

◆ save_checkpoint_components() [1/2]

◆ save_checkpoint_components() [2/2]

◆ step() [1/2]

◆ step() [2/2]

◆ validate() [1/2]

◆ validate() [2/2]

◆ write() [1/2]

◆ write() [2/2]

◆ write_log()

Member Data Documentation

◆ max_design_change

◆ norm2_design_change