No noise HO

TadaSanar · Apr 14, 2023 · 2c1cc69 · 2c1cc69
1 parent 80bcd23
commit 2c1cc69
Show file tree

Hide file tree

Showing 3 changed files with 43 additions and 26 deletions.
diff --git a/hper_bo.py b/hper_bo.py
@@ -570,7 +570,13 @@ def determine_data_fusion_points(data_fusion_XZ_accum,
         # Pick new points for which the surrogate model has a high gradient, no
         # matter if there is an earlier data fusion point nearby.
         new_df_points_x_g = x_next[k][grad_max_s_next > gradient_limit]
-
+
+        print('\nGradients of points surviving gradient limit:\n', 
+              grad_max_s_next[grad_max_s_next > gradient_limit])
+
+        print('\nGradients of points not surviving gradient limit:\n', 
+              grad_max_s_next[grad_max_s_next < gradient_limit])
+
         # THIS OPTION IS NOT IN USE.
         # Pick new points for which the surrogate model does not have
         # a high gradient but they are located far away from the
@@ -691,7 +697,8 @@ def bo_sim_target(bo_ground_truth_model_path='./Source_data/C2a_GPR_model_with_u
                   init_points=None, batch_size=1,
                   acquisition_function='EI', acq_fun_params=None,
                   df_data_coll_params=None, no_plots=False,
-                  results_folder='./Results/'):
+                  results_folder='./Results/', noise_df = False, 
+                  noise_target = False):
     '''
     Simulates a Bayesian Optimization cycle using the Gaussian Process
     Regression model given in bo_ground_truth_model_path as the ground truth for
@@ -878,9 +885,9 @@ def bo_sim_target(bo_ground_truth_model_path='./Source_data/C2a_GPR_model_with_u
 
         if df_data_coll_params is not None:
 
-            #print('Data fusion requests round ' +
-            #      str(k-1) + ':', data_fusion_x_next[k-1])
-            #print('Target requests round ' + str(k-1) + ':', x_next[k-1])
+            print('Data fusion requests round ' +
+                  str(k-1) + ':', data_fusion_x_next[k-1])
+            print('Target requests round ' + str(k-1) + ':', x_next[k-1])
 
             # Do data fusion.
             if data_fusion_XZ_rounds[k] is None:
@@ -890,7 +897,7 @@ def bo_sim_target(bo_ground_truth_model_path='./Source_data/C2a_GPR_model_with_u
                 data_fusion_XZ_rounds = query_data_fusion_XZ_this_round(
                     data_fusion_XZ_rounds, data_fusion_x_next,
                     df_data_coll_params, acq_fun_params, k,
-                    data_fusion_gt_model, noise=False)
+                    data_fusion_gt_model, noise = noise_df)
                 # TO DO: Add noise to the simulated prediction! Confirm proper
                 # scaling in GP once more. Divide the function so that
                 # experimental treatment is fully before BO loop  and simulated
@@ -918,12 +925,10 @@ def bo_sim_target(bo_ground_truth_model_path='./Source_data/C2a_GPR_model_with_u
             # and update X_rounds, Y_rounds, X_accum, Y_accum in place.
             X_rounds, Y_rounds, X_accum, Y_accum = query_target_data_from_model(
                 k, X_rounds, Y_rounds, X_accum, Y_accum, init_points, x_next,
-                stability_model, rounds, materials, noise=False)
-
-            #print('Target data round ' + str(k) + ':', X_accum[k], Y_accum[k])
-
-        #print('X_step and Y_step for round ' + str(k) + ':', X_step[k], Y_step[k])
+                stability_model, rounds, materials, noise = noise_target)
 
+            print('Target data round ' + str(k) + ':', X_accum[k], Y_accum[k])
+
         # Define and fit BO object.
         # f=None because this code will be adapted in future for experimental
         # BO cycle.
@@ -937,7 +942,11 @@ def bo_sim_target(bo_ground_truth_model_path='./Source_data/C2a_GPR_model_with_u
                                                             evaluator_type='local_penalization',
                                                             batch_size=batch_size,
                                                             acquisition_jitter=jitter,
-                                                            acq_fun_params=acq_fun_params
+                                                            acq_fun_params=acq_fun_params,
+                                                            noise_var = 0.1,
+                                                            optimize_restarts = 10,
+                                                            max_iters = 10000,
+                                                            exact_feval = (not noise_target)
                                                             )
 
         # Suggest next points (samples to prepare).
@@ -960,7 +969,7 @@ def bo_sim_target(bo_ground_truth_model_path='./Source_data/C2a_GPR_model_with_u
         current_surrogate_model_params = {'lengthscale': lengthscales[k],
                                           'variance': variances[k],
                                           'max_gradient': max_gradients[k]}
-
+        
         if df_data_coll_params is not None:
 
             # Do data fusion.

diff --git a/hper_repetitions_parallel.py b/hper_repetitions_parallel.py
@@ -133,9 +133,9 @@ def modify_filename_nreps(filename, new_value, param_to_modify_str = '_nreps'):
 
 def repeated_tests(m):    
 
-    c_eig = [0.1, 1, 2, 5] # Expected information gain.
+    c_eig = [0.05, 0.1, 1, 2, 5, 8, 10] # Expected information gain.
     c_exclz = [1, 5, 10, 20] # Size of the exclusion zone in percentage points (max. 100)
-    c_g = [0.063, 0.253, 0.674, 1.282, 1.96]#list(cg(np.array([0, 0.01, 0.05, 0.2, 0.5, 0.8, 0.95, 0.99, 1]))) # Gradient limit. 0.05#, 0.07, 0.1, 0.2, 0.5, 0.75
+    c_g = [0, 0.063, 0.253, 0.674, 1.282, 1.96]#list(cg(np.array([0, 0.01, 0.05, 0.2, 0.5, 0.8, 0.95, 0.99, 1]))) # Gradient limit. 0.05#, 0.07, 0.1, 0.2, 0.5, 0.75
 
     hyperparams_eig = []
     hyperparams_exclz = []
@@ -148,11 +148,11 @@ def repeated_tests(m):
 
             hyperparams_eig.append((c_g[i], c_eig[j]))
 
-    folder = './Results/20230413-jitter01-noiseless/'
+    folder = './Results/20230414-jitter01-noiseless-HO/'
     ground_truth = [0.17, 0.03, 0.80]  # From C2a paper
 
     bo_params = {'n_repetitions': 25,
-                 'n_rounds': 100,
+                 'n_rounds': 15,
                  'n_init': 2,
                  'batch_size': 1,
                  'materials': ['CsPbI', 'MAPbI', 'FAPbI']
@@ -162,8 +162,11 @@ def repeated_tests(m):
     fetch_old_results = False
     # Give False if you don't want to save the figures.
     save_figs = True
+    # Choose if noisy queries are being used or exact.
+    noise_df= False
+    noise_target = False
 
-    if (m > 1): # TO DO change
+    if (m > -1):
 
         if m == 0:
 
@@ -298,7 +301,8 @@ def repeated_tests(m):
                     acquisition_function = acquisition_function,
                     acq_fun_params = afp,
                     df_data_coll_params = ddcp,
-                    no_plots=no_plots, results_folder = triangle_folder)
+                    no_plots=no_plots, results_folder = triangle_folder,
+                    noise_df = noise_df, noise_target = noise_target)
 
                 # All ouputs saved only from the first two repetitions to save
                 # disk space.
@@ -486,7 +490,7 @@ def repeated_tests(m):
 
     print(os.getcwd())
 
-    m_total = 42
+    m_total = 68
 
     ###############################################################################
     # get number of cpus available to job

diff --git a/hper_repetitions_parallel_temp_noiseless.py b/hper_repetitions_parallel_temp_noiseless.py
@@ -133,9 +133,9 @@ def modify_filename_nreps(filename, new_value, param_to_modify_str = '_nreps'):
 
 def repeated_tests(m):    
 
-    c_eig = [0.1, 1, 2, 5] # Expected information gain.
+    c_eig = [0.05, 0.1, 1, 2, 5, 8, 10] # Expected information gain.
     c_exclz = [1, 5, 10, 20] # Size of the exclusion zone in percentage points (max. 100)
-    c_g = [0.063, 0.253, 0.674, 1.282, 1.96]#list(cg(np.array([0, 0.01, 0.05, 0.2, 0.5, 0.8, 0.95, 0.99, 1]))) # Gradient limit. 0.05#, 0.07, 0.1, 0.2, 0.5, 0.75
+    c_g = [0, 0.063, 0.253, 0.674, 1.282, 1.96]#list(cg(np.array([0, 0.01, 0.05, 0.2, 0.5, 0.8, 0.95, 0.99, 1]))) # Gradient limit. 0.05#, 0.07, 0.1, 0.2, 0.5, 0.75
 
     hyperparams_eig = []
     hyperparams_exclz = []
@@ -148,11 +148,11 @@ def repeated_tests(m):
 
             hyperparams_eig.append((c_g[i], c_eig[j]))
 
-    folder = './Results/20230413-jitter001-noiseless/'
+    folder = './Results/20230414-jitter01-noiseless-HO/'
     ground_truth = [0.17, 0.03, 0.80]  # From C2a paper
 
     bo_params = {'n_repetitions': 25,
-                 'n_rounds': 100,
+                 'n_rounds': 15,
                  'n_init': 2,
                  'batch_size': 1,
                  'materials': ['CsPbI', 'MAPbI', 'FAPbI']
@@ -162,6 +162,9 @@ def repeated_tests(m):
     fetch_old_results = False
     # Give False if you don't want to save the figures.
     save_figs = True
+    # Choose if noisy queries are being used or exact.
+    noise_df= False
+    noise_target = False
 
     if (m > -1):
 
@@ -298,7 +301,8 @@ def repeated_tests(m):
                     acquisition_function = acquisition_function,
                     acq_fun_params = afp,
                     df_data_coll_params = ddcp,
-                    no_plots=no_plots, results_folder = triangle_folder)
+                    no_plots=no_plots, results_folder = triangle_folder,
+                    noise_df = noise_df, noise_target = noise_target)
 
                 # All ouputs saved only from the first two repetitions to save
                 # disk space.
@@ -486,7 +490,7 @@ def repeated_tests(m):
 
     print(os.getcwd())
 
-    m_total = 42
+    m_total = 68
 
     ###############################################################################
     # get number of cpus available to job