🚧 Renaming attributes to maintain coherence.

digneapy/archives/_grid_archive.py

@@ -38,7 +38,7 @@ def __init__(
         self,
         dimensions: Sequence[int],
         ranges: Sequence[Tuple[float, float]],
-        attribute: str = "features",
+        descriptor: str = "features",
         instances: Optional[Iterable[Instance]] = None,
         eps: float = 1e-6,
         dtype=np.float64,
@@ -57,7 +57,7 @@ def __init__(
             :math:`[-2,2]` (inclusive). ``ranges`` should be the same length as
             ``dims``.
             instances (Optional[Iterable[Instance]], optional): Instances to pre-initialise the archive. Defaults to None.
-            attribute: str = Attribute of the Instances to compute the diversity.
+            descriptor: str = Descriptor of the Instances to compute the diversity.
             eps (float, optional): Due to floating point precision errors, we add a small
             epsilon when computing the archive indices in the :meth:`index_of`
             method -- refer to the implementation `here. Defaults to 1e-6.
@@ -76,15 +76,15 @@ def __init__(
             )
 
         self._dimensions = np.asarray(dimensions)
-        self._inst_attrs = attribute
-        if attribute not in descriptor_strategies:
-            msg = f"describe_by {attribute} not available in {self.__class__.__name__}.__init__. Set to features by default"
+        self._inst_attrs = descriptor
+        if descriptor not in descriptor_strategies:
+            msg = f"describe_by {descriptor} not available in {self.__class__.__name__}.__init__. Set to features by default"
             print(msg)
             self._inst_attrs = "features"
             self._descriptor_strategy = descriptor_strategies["features"]
         else:
-            self._inst_attrs = attribute
-            self._descriptor_strategy = descriptor_strategies[attribute]
+            self._inst_attrs = descriptor
+            self._descriptor_strategy = descriptor_strategies[descriptor]
 
         ranges = list(zip(*ranges))
         self._lower_bounds = np.array(ranges[0], dtype=dtype)

digneapy/generators/_map_elites_gen.py

@@ -37,7 +37,7 @@ def __init__(
         archive: GridArchive,
         mutation: mutation.Mutation,
         repetitions: int,
-        strategy: str,
+        descriptor: str,
         performance_function: PerformanceFn = default_performance_metric,
     ):
         self._domain = domain
@@ -49,17 +49,17 @@ def __init__(
         self._repetitions = repetitions
         self._performance_fn = performance_function
 
-        if strategy not in descriptor_strategies:
-            msg = f"strategy {strategy} not available in {self.__class__.__name__}.__init__. Set to features by default"
+        if descriptor not in descriptor_strategies:
+            msg = f"descriptor {descriptor} not available in {self.__class__.__name__}.__init__. Set to features by default"
             print(msg)
-            strategy = "features"
+            descriptor = "features"
 
-        self._descriptor = strategy
-        match strategy:
+        self._descriptor = descriptor
+        match descriptor:
             case "features":
                 self._descriptor_strategy = self._domain.extract_features
             case _:
-                self._descriptor_strategy = descriptor_strategies[strategy]
+                self._descriptor_strategy = descriptor_strategies[descriptor]
 
         self._stats_fitness = tools.Statistics(key=attrgetter("fitness"))
         self._stats_fitness.register("avg", np.mean)

examples/evolve_nn_pytorch.py

@@ -46,7 +46,7 @@ class NSEval:
     This must be called for each transformed at every generation of the CMA-ES algorithm.
     """
 
-    def __init__(self, features_info, resolution: int = 20):
+    def __init__(self, features_info, resolution: int = 10):
         self.resolution = resolution
         self.features_info = features_info
         self.kp_domain = KPDomain(dimension=50, capacity_approach="percentage")
@@ -106,7 +106,7 @@ def __call__(self, transformer: TorchNN):
 
 
 def main():
-    R = 20  # Resolution/Number of bins for each of the 8 features
+    R = 10  # Resolution/Number of bins for each of the 8 features
     dimension = 46  # Number of weights of the NN for KP
     nn = TorchNN(
         name="NN_transformer_kp_domain.keras",
@@ -115,16 +115,26 @@ def main():
         output_size=2,
     )
 
-    # # KP Features information extracted from previously generated instances
+    # KP Features information extracted from previously generated instances
+    # features_info = [
+    #     (711, 30000),
+    #     (890, 1000),
+    #     (860, 1000.0),
+    #     (1.0, 200),
+    #     (1.0, 230.0),
+    #     (0.10, 12.0),
+    #     (400, 610),
+    #     (240, 330),
+    # ]
     features_info = [
-        (711, 30000),
-        (890, 1000),
-        (860, 1000.0),
-        (1.0, 200),
-        (1.0, 230.0),
-        (0.10, 12.0),
-        (400, 610),
-        (240, 330),
+        (1, 30000),  # Capacity
+        (1, 1000),  # Max Profit
+        (1, 1000),  # Max Weight
+        (1, 1000),  # Min Profit
+        (1, 1000),  # Min Weight
+        (0.0, 100.0),  # Average Efficiency
+        (0.0, 1000),  # Mean of vars
+        (0.0, 500.0),  # Std
     ]
     # NSEval is the evaluation/fitness function used to measure the NNs in CMA-Es
     ns_eval = NSEval(features_info, resolution=R)

examples/knapsack_domain_map_elites.py

@@ -40,7 +40,7 @@ def map_elites_knapsack():
         initial_pop_size=10,
         mutation=uniform_one_mutation,
         generations=1000,
-        strategy="features",
+        descriptor="features",
         repetitions=1,
     )
     assert archive.filled_cells == len(archive.instances)

tests/generators/test_map_elites.py

@@ -58,7 +58,7 @@ def test_map_elites_domain(
         initial_pop_size=5,
         mutation=uniform_one_mutation,
         generations=1000,
-        strategy="features",
+        descriptor="features",
         repetitions=1,
     )
     archive = map_elites()