Update genetic_algorithm.py to remove ordinal variables

jandalee · Aug 30, 2023 · 2bccb4e · 2bccb4e
1 parent 3431140
commit 2bccb4e
Showing 1 changed file with 11 additions and 9 deletions.
diff --git a/MCBO/mcbo/optimizers/non_bo/genetic_algorithm.py b/MCBO/mcbo/optimizers/non_bo/genetic_algorithm.py
@@ -294,8 +294,8 @@ def __init__(self,
                  dtype: torch.dtype = torch.float64,
                  ):
 
-        assert search_space.num_nominal + search_space.num_ordinal == search_space.num_dims, \
-            'Genetic Algorithm currently supports only nominal and ordinal variables'
+        assert search_space.num_nominal == search_space.num_dims, \
+            'Genetic Algorithm currently supports only nominal variables'
 
         super(CategoricalGeneticAlgorithm, self).__init__(
             search_space=search_space, dtype=dtype, input_constraints=input_constraints
@@ -336,11 +336,11 @@ def __init__(self,
         if self.tr_manager is not None:
             self.x_queue.iloc[0:1] = self.search_space.inverse_transform(self.tr_center.unsqueeze(0))
 
-        self.map_to_canonical = self.search_space.nominal_dims + self.search_space.ordinal_dims
+        self.map_to_canonical = self.search_space.nominal_dims
         self.map_to_original = [self.map_to_canonical.index(i) for i in range(len(self.map_to_canonical))]
 
-        self.lb = self.search_space.nominal_lb + self.search_space.ordinal_lb
-        self.ub = self.search_space.nominal_ub + self.search_space.ordinal_ub
+        self.lb = self.search_space.nominal_lb
+        self.ub = self.search_space.nominal_ub
 
     def get_tr_point_sampler(self) -> Callable[[int], pd.DataFrame]:
         """
@@ -422,7 +422,8 @@ def method_suggest(self, n_suggestions: int = 1) -> pd.DataFrame:
         if n_remaining and len(self.x_queue):
             n = min(n_remaining, len(self.x_queue))
             x_next.iloc[idx: idx + n] = self.x_queue.iloc[idx: idx + n]
-            self.x_queue = self.x_queue.drop(self.x_queue.index[[i for i in range(idx, idx + n)]]).reset_index(drop=True)
+            self.x_queue = self.x_queue.drop(self.x_queue.index[[i for i in range(idx, idx + n)]]).reset_index(
+                drop=True)
 
             idx += n
             n_remaining -= n
@@ -432,7 +433,8 @@ def method_suggest(self, n_suggestions: int = 1) -> pd.DataFrame:
 
             n = min(n_remaining, len(self.x_queue))
             x_next.iloc[idx: idx + n] = self.x_queue.iloc[idx: idx + n]
-            self.x_queue = self.x_queue.drop(self.x_queue.index[[i for i in range(idx, idx + n)]]).reset_index(drop=True)
+            self.x_queue = self.x_queue.drop(self.x_queue.index[[i for i in range(idx, idx + n)]]).reset_index(
+                drop=True)
 
             idx += n
             n_remaining -= n
@@ -615,7 +617,7 @@ def point_sampler(n_points: int) -> pd.DataFrame:
         return
 
     def _crossover(self, x1: torch.Tensor, x2: torch.Tensor) -> (torch.Tensor, torch.Tensor):
-        assert self.search_space.num_ordinal + self.search_space.num_nominal == self.search_space.num_dims, \
+        assert self.search_space.num_nominal == self.search_space.num_dims, \
             'Current crossover can\'t handle permutations'
 
         x1_ = x1.clone()
@@ -655,7 +657,7 @@ def _crossover(self, x1: torch.Tensor, x2: torch.Tensor) -> (torch.Tensor, torch
         return x1_, x2_
 
     def _mutate(self, x: torch.Tensor) -> torch.Tensor:
-        assert self.search_space.num_ordinal + self.search_space.num_nominal == self.search_space.num_dims, \
+        assert self.search_space.num_nominal == self.search_space.num_dims, \
             'Current mutate can\'t handle permutations'
         assert x.ndim == 2, (x.shape, self.map_to_canonical)
         x_ = x.clone()[:, self.map_to_canonical]