lin_evol_gen_reduced_data_primal_dual.m File Reference

reduced_data = lin_evol_gen_reduced_data_primal_dual(model, detailed_data) More...

Go to the source code of this file.

Functions
function reduced_data =	lin_evol_gen_reduced_data_primal_dual (model, detailed_data)
	reduced_data = lin_evol_gen_reduced_data_primal_dual(model, detailed_data) More...

Detailed Description

reduced_data = lin_evol_gen_reduced_data_primal_dual(model, detailed_data)

Definition in file lin_evol_gen_reduced_data_primal_dual.m.

Function Documentation

function reduced_data = lin_evol_gen_reduced_data_primal_dual	(		model,
			detailed_data
	)

reduced_data = lin_evol_gen_reduced_data_primal_dual(model, detailed_data)

method which produces reduced_data, which is the data, that will be passed to an online-algorithm. Therefore, no quantities dependent on the high-dimension H may be included here. Neither may online-data include parameter-dependent mu-quantities. So no complete grid or detailed solutions or reduced basis vectors may be stored here. So online data is produced in the offline stage, but may be used in online-stages. So the computation time may depend on H, but the results may not depend on this complexity.

allowed dependency of generated data: Nmax not allowed dependency of data: H allowed dependency of computation: Nmax, H Unknown at this stage: mu, N

Bernard Haasdonk 16.5.2007

extended to primal/dual formulaton by Dominik Garmatter

the following cases are possible

- model.want_improved_output = 0 and model.want_dual = 1 or 0

=> lin_evol_gen_reduced_data_primal_dual(model, detailed_data) with detailed_data including the primal or dual detailed data. Simply produced reduced_data for a primal or a dual problem. In the dual case a functional has to be specified via model.name_output_functional.

- model.want_improved_output = 1 and model.want_dual = 0

=> lin_evol_gen_reduced_data_primal_dual(model, detailed_data) with detailed_data including both the primal and dual detailed_data! See (and use) lin_evol_primal_dual_gen_detailed_data, lin_evol_primal_dual_gen_reduced_data and lin_evol_primal_dual_rb_simulation for this case.

In all the cases it is possible to enable the SCM via model.use_scm = 1 to compute scm_offline_data for a lower bound to the infsup or a coercivity constant in the rb_simulation. Only makes sense in the above second case or if you set model.error_norm = energy for the future rb_simulation. See scm_offline for a description on the required scm quantities.

special Note: in the european_option_pricing model a functional named theta includes a partial timederivative and therefore produces a slightly different dual problem. Therefore the theta-case is sometimes treated differently. Dont care about this case if you dont use the european_option_pricing model.

Parameters

model	model
detailed_data	detailed data

Return values

reduced_data

reduced data

Required fields of model:

• want_dual —  want dual
• name_output_functional —  name output functional
• use_scm —  use scm
• want_improved_output —  want improved output
• get_rb_size —  get rb size
• operators_output —  operators output

Required fields of detailed_data:

• RB —  RB

Generated fields of reduced_data:

• N —  N
• a0 —  a0
• s_l2norm —  s l2norm
• s_RB —  s RB

Definition at line 17 of file lin_evol_gen_reduced_data_primal_dual.m.

Here is the caller graph for this function: