update warmstart

pyomo/contrib/mindtpy/util.py

@@ -38,7 +38,8 @@
 from pyomo.solvers.plugins.solvers.gurobi_persistent import GurobiPersistent
 import math
 from itertools import permutations, product
-import pickle 
+import pickle
+import numpy as np
 
 pyomo_nlp = attempt_import('pyomo.contrib.pynumero.interfaces.pyomo_nlp')[0]
 numpy = attempt_import('numpy')[0]
@@ -971,6 +972,8 @@ def generate_norm_constraint(fp_nlp_model, mip_model, config):
 def customized_initialize(mod):
     with open("rotary_bed_unit_FIM", 'rb') as f:
         unit_fim_list = pickle.load(f)
+
+    new_fim = np.zeros((5,5))
 
     ### compute FIM 
     def eval_fim2(m, a, b):
@@ -985,40 +988,39 @@ def eval_fim2(m, a, b):
             for i in m.n_responses:
                 for j in m.n_responses:
                     if j>=i:
-                        summi += m.cov_y[i,j]*unit_fim_list[i*m.num_measure_dynamic_flatten+j][a][b]
+                        summi += m.cov_y[i,j].value*unit_fim_list[i*m.num_measure_dynamic_flatten+j][a][b]
                     else:
-                        summi += m.cov_y[j,i]*unit_fim_list[i*m.num_measure_dynamic_flatten+j][a][b]
-
-
+                        summi += m.cov_y[j,i].value*unit_fim_list[i*m.num_measure_dynamic_flatten+j][a][b]
             return summi
         else:
             summi = 0 
             for i in m.n_responses:
                 for j in m.n_responses:
                     if j>=i:
-                        summi += m.cov_y[i,j]*unit_fim_list[i*m.num_measure_dynamic_flatten+j][b][a]
+                        summi += m.cov_y[i,j].value*unit_fim_list[i*m.num_measure_dynamic_flatten+j][b][a]
                     else:
-                        summi += m.cov_y[j,i]*unit_fim_list[i*m.num_measure_dynamic_flatten+j][b][a]
-
-
+                        summi += m.cov_y[j,i].value*unit_fim_list[i*m.num_measure_dynamic_flatten+j][b][a]
             return summi
-
-    #mod.dynamic_initial_FIM = pyo.Expression(mod.DimFIM_half, rule=eval_fim2)
-
+
     for a in mod.DimFIM:
         for b in mod.DimFIM:
             if a<=b:
                 dynamic_initial_element = eval_fim2(mod, a,b)
                 mod.TotalFIM[a,b].value = dynamic_initial_element
-                #new_fim[a,b] = dynamic_initial_element
+                new_fim[a,b] = dynamic_initial_element
+                new_fim[b,a] = dynamic_initial_element
 
                 ### update greybox
-                #grey_box_name = "ele_"+str(a)+"_"+str(b)
-                #mod.my_block.egb.inputs[grey_box_name].value = dynamic_initial_element
+                grey_box_name = "ele_"+str(a)+"_"+str(b)
+                mod.my_block.egb.inputs[grey_box_name].value = dynamic_initial_element
+
+    # initialize determinant 
+    _, det = np.linalg.slogdet(new_fim)
+    mod.my_block.egb.outputs["log_det"] = det 
 
     # manual and dynamic install initial 
     def total_dynamic_exp(m):
-        return sum(m.cov_y[i,i] for i in range(m.n_static_measurements, m.num_measure_dynamic_flatten))
+        return sum(m.cov_y[i,i].value for i in range(m.n_static_measurements, m.num_measure_dynamic_flatten))
 
     total_dynamic_initial = total_dynamic_exp(mod)
     mod.total_number_dynamic_measurements.value = total_dynamic_initial 
@@ -1028,10 +1030,9 @@ def cost_compute(m):
         """Compute cost
         cost = static-cost measurement cost + dynamic-cost measurement installation cost + dynamic-cost meausrement timepoint cost 
         """
-        return sum(m.cov_y[i,i]*m.cost_list[i] for i in m.n_responses)+sum(m.if_install_dynamic[j]*m.dynamic_install_cost[j-m.n_static_measurements] for j in m.DimDynamic)
+        return sum(m.cov_y[i,i].value*m.cost_list[i] for i in m.n_responses)+sum(m.if_install_dynamic[j].value*m.dynamic_install_cost[j-m.n_static_measurements] for j in m.DimDynamic)
 
     cost_init = cost_compute(mod)
     mod.cost.value = cost_init
 
     return mod
-