preparation for batch optimization

env/optimal_parameters_b_chlorine_cdrec.toml

@@ -0,0 +1,4 @@
+[cdrec]
+rank = "8"
+epsilon = 0.00018675637037515554
+iteration = "1000"

env/optimal_parameters_b_chlorine_stmvl.toml

@@ -0,0 +1,4 @@
+[stmvl]
+window_size = "2"
+gamma = 0.999999
+alpha = "6"

imputegap/manager/utils.py

@@ -30,7 +30,7 @@ def get_save_path_asset():
     return filepath
 
 
-def load_parameters(query: str = "default", algorithm: str = "cdrec"):
+def load_parameters(query: str = "default", algorithm: str = "cdrec", dataset: str = "chlorine", optimizer: str="b"):
     """
     Load default values of algorithms
 
@@ -44,7 +44,7 @@ def load_parameters(query: str = "default", algorithm: str = "cdrec"):
     if query == "default":
         filepath = "../env/default_values.toml"
     elif query == "optimal":
-        filepath = "../env/optimal_parameters_"+str(algorithm)+".toml"
+        filepath = "../env/optimal_parameters_"+str(optimizer)+"_"+str(dataset)+"_"+str(algorithm)+".toml"
     else:
         print("Query not found for this function ('optimal' or 'default')")
 

imputegap/optimization/bayesian_optimization.py

@@ -15,15 +15,15 @@
 
 class Optimization:
 
-    def save_optimization(optimal_params, algorithm="cdrec", file_name=None):
+    def save_optimization(optimal_params, algorithm="cdrec", dataset="", optimizer="b", file_name=None):
         """
         Save the optimization parameters to a TOML file to use later without recomputing.
 
         :param optimal_params: dictionary of the optimal parameters.
         :param file_name: name of the TOML file to save the results. Default is 'optimization_results.toml'.
         """
         if file_name is None:
-            file_name = "../env/optimal_parameters_" + str(algorithm) + ".toml"
+            file_name = "../env/optimal_parameters_" + str(optimizer) + "_" + str(dataset) + "_" + str(algorithm) + ".toml"
 
         if not os.path.exists(file_name):
             file_name = file_name[1:]
@@ -40,7 +40,7 @@ def save_optimization(optimal_params, algorithm="cdrec", file_name=None):
     class Bayesian:
 
         def bayesian_optimization(ground_truth, contamination, selected_metrics=["RMSE"], algorithm="cdrec",
-                                  n_calls=10, n_random_starts=50, acq_func='gp_hedge'):
+                                  n_calls=100, n_random_starts=50, acq_func='gp_hedge'):
             """
             Conduct the Bayesian optimization hyperparameter optimization.
 

imputegap/runner_optimization.py

@@ -23,25 +23,27 @@ def check_block_size(filename):
 if __name__ == '__main__':
 
     display_title()
+    datasets = ["bafu", "chlorine", "climate", "drift", "egg", "meteo", "test-large"]
 
-    filename = "eeg"
-    file_path = os.path.join("./dataset/", filename + ".txt")
-    gap = TimeSeries(data=file_path)
+    for filename in datasets :
+        file_path = os.path.join("./dataset/", filename + ".txt")
+        gap = TimeSeries(data=file_path)
 
-    block_size, plot_limit = check_block_size(filename)
+        block_size, plot_limit = check_block_size(filename)
 
-    gap.print(limitation=5)
-    gap.plot(title="test", save_path="assets", limitation=6, display=False)
+        gap.print(limitation=5)
+        gap.plot(title="test", save_path="assets", limitation=6, display=False)
 
-    gap.ts_contaminate = Contamination.scenario_mcar(ts=gap.ts, series_impacted=0.4, missing_rate=0.4, block_size=block_size, protection=0.1, use_seed=True, seed=42)
-    gap.print()
-    gap.plot(ts_type="contamination", title="test", save_path="assets", limitation=3, display=False)
+        gap.ts_contaminate = Contamination.scenario_mcar(ts=gap.ts, series_impacted=0.4, missing_rate=0.4, block_size=block_size, protection=0.1, use_seed=True, seed=42)
+        gap.print()
+        gap.plot(ts_type="contamination", title="test", save_path="assets", limitation=3, display=False)
 
-    for algo in ["cdrec", "stmvl", "iim", "mrnn"]:
-        print("RUN OPTIMIZATION FOR : ", algo, "...")
-        optimal_params, yi = Optimization.Bayesian.bayesian_optimization(ground_truth=gap.ts, contamination=gap.ts_contaminate, algorithm=algo)
-        print("\nOptical Params : ", optimal_params)
-        print("\nyi : ", yi, "\n")
-        Optimization.save_optimization(optimal_params=optimal_params, algorithm=algo)
-        print("\n", "_"*95, "end")
+        for algo in ["cdrec", "stmvl", "iim", "mrnn"]:
+            print("RUN OPTIMIZATION FOR : ", algo, "... with ", filename, "...")
+            optimal_params, yi = Optimization.Bayesian.bayesian_optimization(ground_truth=gap.ts, contamination=gap.ts_contaminate, algorithm=algo)
+            print("\nOptical Params : ", optimal_params)
+            print("\nyi : ", yi, "\n")
+            Optimization.save_optimization(optimal_params=optimal_params, algorithm=algo, dataset=filename, optimizer="b")
+            print("\n", "_"*95, "end")
+        print("\n", "_" * 95, "end")
     print("\n", "_" * 95, "end")

tests/test_opti_bayesian_cdrec.py

@@ -15,9 +15,10 @@ def test_optimization_bayesian_cdrec(self):
         """
         the goal is to test if only the simple optimization with cdrec has the expected outcome
         """
-        gap = TimeSeries(utils.get_file_path_dataset("chlorine"))
-
         algorithm = "cdrec"
+        dataset = "chlorine"
+
+        gap = TimeSeries(utils.get_file_path_dataset(dataset))
 
         ts_contaminated = Contamination.scenario_mcar(ts=gap.ts, series_impacted=0.4, missing_rate=0.4, block_size=2, protection=0.1, use_seed=True, seed=42)
 
@@ -32,7 +33,5 @@ def test_optimization_bayesian_cdrec(self):
         _, metrics_optimal = Imputation.MR.cdrec(ground_truth=gap.ts, contamination=ts_contaminated, params=params_optimal)
         _, metrics_default = Imputation.MR.cdrec(ground_truth=gap.ts, contamination=ts_contaminated, params=params)
 
-        Optimization.save_optimization(optimal_params=optimal_params, algorithm=algorithm+"_test")
-
         self.assertTrue(metrics_optimal["RMSE"] < metrics_default["RMSE"], f"Expected {metrics_optimal['RMSE']} > {metrics_default['RMSE']}")
         self.assertTrue(yi > 0, True)

tests/test_opti_bayesian_iim.py

@@ -36,7 +36,5 @@ def test_optimization_bayesian_iim(self):
         _, metrics_optimal = Imputation.Regression.iim_imputation(ground_truth=gap.ts, contamination=ts_contaminated, params=params_optimal)
         _, metrics_default = Imputation.Regression.iim_imputation(ground_truth=gap.ts, contamination=ts_contaminated, params=params)
 
-        Optimization.save_optimization(optimal_params=optimal_params, algorithm=algorithm+"_test")
-
         self.assertTrue(metrics_optimal["RMSE"] < metrics_default["RMSE"], f"Expected {metrics_optimal['RMSE']} > {metrics_default['RMSE']}")
         self.assertTrue(yi > 0, True)