cloning initial best solucion to current solution to avoid aliasing

between the best and the current solutions

Author: Laouen

thoi/heuristics/simulated_annealing.py

@@ -111,15 +111,15 @@ def simulated_annealing(X: np.ndarray,
 
     # Best solution found
     # |batch_size| x |order|
-    best_solution = current_solution
+    best_solution = current_solution.clone()
     # |batch_size|
-    best_energy = current_energy
+    best_energy = current_energy.clone()
 
     no_progress_count = 0
     pbar = trange(max_iterations, leave=False)
     for _ in pbar:
 
-        pbar.set_description(f'mean({metric.upper()}) = {(1 if largest else -1) * best_energy.mean()}')
+        pbar.set_description(f'mean({metric.upper()}) = {(1 if largest else -1) * best_energy.mean()} - ES: {no_progress_count}')
 
         # Generate new solution by modifying the current solution
         # |batch_size| x |order|

thoi/heuristics/simulated_annealing_multi_order.py

@@ -122,15 +122,15 @@ def simulated_annealing_multi_order(X: np.ndarray,
 
     # Best solution found
     # |batch_size| x |order|
-    best_solution = current_solution
+    best_solution = current_solution.clone()
     # |batch_size|
-    best_energy = current_energy
+    best_energy = current_energy.clone()
 
     no_progress_count = 0
-    pbar = trange(max_iterations, leave=False)
+    pbar = trange(max_iterations, leave=True)
     for _ in pbar:
 
-        pbar.set_description(f'mean({metric.upper()}) = {(1 if largest else -1) * best_energy.mean()}')
+        pbar.set_description(f'mean({metric.upper()}) = {(1 if largest else -1) * best_energy.mean()} - ES: {no_progress_count}')
 
 
         # |batch_size|