#@ p*: Label=FastTest
#@ p0: MPIProcs=2

# DAKOTA INPUT FILE - dakota_sbo_textbook.in

# Demonstrates the use of approximation models and a trust region
# optimization environment in performing constrained minimization on
# the textbook test function.

# These tests exercise the "correction" capabilities of the surrogate
# based optimization environment. Here, the options are "none," "additive
# zeroth_order," "multiplicative zeroth_order," "additive first_order,"
# and "multiplicative first_order."  The "none" case is the default
# setting which is invoked if the "correction" keyword is omitted.

# Test 8 mirrors 5, but with the new GP module

environment,
	method_pointer = 'SBLO'

method,
	id_method = 'SBLO'
	surrogate_based_local
	model_pointer = 'SURROGATE'
	approx_method_pointer = 'NLP'
	max_iterations = 50
	trust_region
	  initial_size = 0.10
	  contraction_factor = 0.5
	  expansion_factor   = 1.50
	  soft_convergence   = 1    # slowed convergence

method,
	id_method = 'NLP'
	conmin_mfd
#	optpp_cg
#	npsol
#	dot_bfgs
#	dot_fr
#	dot_mmfd
	  max_iterations = 50
	  convergence_tolerance = 1e-4

model,
	id_model = 'SURROGATE'
	responses_pointer = 'SURROGATE_RESP'
	surrogate global
	  dace_method_pointer = 'SAMPLING'
#	  reuse_samples region
#	  correction additive       zeroth_order #s1,#s5,#s8,#s9
#	  correction multiplicative zeroth_order #s2
#	  correction additive       first_order	 #s3,#s6,#s7,#p0
#	  correction multiplicative first_order	 #s4
#	  use_derivatives			 #s6,#s7
#	  neural_network
	  polynomial quadratic			 #s0,#s1,#s2,#s3,#s4,#p0	
#	  gaussian_process surfpack		 #s5,#s6,#s7
#	  gaussian_process surfpack correlation_lengths =  0.707106781186547 0.707106781186547
#   experimental_gaussian_process		 #s8
#     find_nugget 1 #s8
#     trend quadratic #s8
# Emulate s1 with new polynomial model
#         experimental_polynomial    	 	 #s9
#           basis_order 2			 #s9
#	  mars
# 	surrogate local taylor_series
#	  actual_model_pointer = 'TRUTH'

variables,
	continuous_design = 2
	  initial_point    2.0    2.0
	  upper_bounds     5.8    2.9
	  lower_bounds     0.5   -2.9
	  descriptors      'x1'   'x2'

responses,
	id_responses = 'SURROGATE_RESP'
	objective_functions = 1
	nonlinear_inequality_constraints = 2
#	analytic_gradients			
#	no_gradients				
  	numerical_gradients			
  	  method_source dakota			
  	  interval_type forward			
  	  fd_gradient_step_size = .0001		
#	analytic_hessians
	no_hessians

###############################################
# SAMPLING method specifications for building #
# surrogate function(s)			      #
###############################################
method,
	id_method = 'SAMPLING'
	model_pointer = 'TRUTH'
#	dace central_composite
#	dace box_behnken
	dace lhs
	  seed = 12345
	  samples = 10
#	dace oas seed = 5
#	  samples = 49 symbols = 7

model,
	id_model = 'TRUTH'
	single
	  interface_pointer = 'TRUE_FN'
	  responses_pointer = 'TRUE_RESP'

interface,
	direct
	id_interface = 'TRUE_FN'
 	  analysis_driver = 'text_book'

responses,
	id_responses = 'TRUE_RESP'
	objective_functions = 1
	nonlinear_inequality_constraints = 2
#	analytic_gradients		#s3,#s4,#s7
	no_gradients		 	#s0,#s1,#s2,#s5,#s8,#s9
#  	numerical_gradients		#s6,#p0
#  	  method_source dakota		#s6,#p0
#  	  interval_type central		#s6,#p0
#  	  fd_gradient_step_size = 1.e-9	#s6,#p0
	no_hessians