Merge pull request #96 from rodrigo-arenas/0.9.0

Add ConstantAdapter

docs/api/schedules.rst

@@ -5,6 +5,7 @@ Schedules
 
 .. autosummary::
    base.BaseAdapter
+   ConstantAdapter
    ExponentialAdapter
    InverseAdapter
    PotentialAdapter
@@ -13,6 +14,10 @@ Schedules
    :members:
    :undoc-members: False
 
+.. autoclass:: ConstantAdapter
+   :members:
+   :undoc-members: False
+
 .. autoclass:: ExponentialAdapter
    :members:
    :undoc-members: False

docs/conf.py

@@ -73,6 +73,9 @@
 
 master_doc = "index"
 
+# Ignore Jupyter notebooks error
+nbsphinx_allow_errors = True
+
 # generate autosummary even if no references
 autosummary_generate = True
 autosummary_imported_members = True

docs/release_notes.rst

@@ -13,6 +13,7 @@ Features:
 * Introducing Adaptive Schedulers to enable adaptive mutation and crossover probabilities;
   currently, supported schedulers are:
 
+  - :class:`~sklearn_genetic.schedules.ConstantAdapter`
   - :class:`~sklearn_genetic.schedules.ExponentialAdapter`
   - :class:`~sklearn_genetic.schedules.InverseAdapter`
   - :class:`~sklearn_genetic.schedules.PotentialAdapter`

docs/tutorials/adapters.rst

@@ -38,7 +38,7 @@ If :math:`p_0 > p_f`, you are performing a decay towards :math:`p_0`.
 
 If :math:`p_0 < p_f`, you are performing an ascend towards :math:`p_f`.
 
-All the adapters :math:`p(t; \alpha)`, for :math:`\alpha \in (0,1)`,
+All the non-constant adapters :math:`p(t; \alpha)`, for :math:`\alpha \in (0,1)`,
 have the following properties:
 
 .. math::
@@ -49,10 +49,25 @@ have the following properties:
 
 The following adapters are available:
 
+* ConstantAdapter
 * ExponentialAdapter
 * InverseAdapter
 * PotentialAdapter
 
+
+ConstantAdapter
+---------------
+
+This adapter is meant to be used internally by the package; when the user doesn't create an adapter but
+instead defines the crossover or mutation probability as a real number, the package will convert it
+to a `ConstantAdapter`, so the library can use the internal API with the same methods in both cases.
+Because of this, its definition is:
+
+.. math::
+
+   p(t; \alpha) = p_0
+
+
 ExponentialAdapter
 ------------------
 

sklearn_genetic/__init__.py

@@ -7,7 +7,12 @@
     LogbookSaver,
 )
 
-from .schedules import ExponentialAdapter, InverseAdapter, PotentialAdapter
+from .schedules import (
+    ConstantAdapter,
+    ExponentialAdapter,
+    InverseAdapter,
+    PotentialAdapter,
+)
 
 from ._version import __version__
 
@@ -18,6 +23,7 @@
     "ConsecutiveStopping",
     "DeltaThreshold",
     "LogbookSaver",
+    "ConstantAdapter",
     "ExponentialAdapter",
     "InverseAdapter",
     "PotentialAdapter",

sklearn_genetic/algorithms.py

@@ -30,11 +30,11 @@ def eaSimple(
     toolbox: A :class:`~deap.base.Toolbox`
         Contains the evolution operators.
 
-    cxpb: float or a Scheduler, default=None
-        The probability of mating two individuals.
+    cxpb: Scheduler, default=None
+        An adaptive scheduler representing the probability of mating two individuals.
 
-    mutpb: float or a Scheduler, default=None
-        The probability of mutating an individual.
+    mutpb: Scheduler, default=None
+        An adaptive scheduler representing the probability that an offspring is produced by mutation.
 
     ngen: int, default=None
         The number of generation.
@@ -124,22 +124,13 @@ def eaSimple(
         print("INFO: Stopping the algorithm")
         return population, logbook, n_gen
 
-    # Begin the generational process
-    mutpb_value = mutpb
-    cxpb_value = cxpb
     for gen in range(1, ngen + 1):
         try:
             # Select the next generation individuals
             offspring = toolbox.select(population, len(population) - hof_size)
 
-            # Check adaptive rates
-            if isinstance(mutpb, BaseAdapter):
-                mutpb_value = mutpb.step()
-            if isinstance(cxpb, BaseAdapter):
-                cxpb_value = cxpb.step()
-
             # Vary the pool of individuals
-            offspring = varAnd(offspring, toolbox, cxpb_value, mutpb_value)
+            offspring = varAnd(offspring, toolbox, cxpb.step(), mutpb.step())
 
             # Evaluate the individuals with an invalid fitness
             invalid_ind = [ind for ind in offspring if not ind.fitness.valid]
@@ -233,11 +224,11 @@ def eaMuPlusLambda(
     lambda\_: int, default=None
         The number of children to produce at each generation.
 
-    cxpb: float or a Scheduler, default=None
+    cxpb: Scheduler, default=None
         The probability that an offspring is produced by crossover.
 
-    mutpb: float or a Scheduler, default=None
-        The probability that an offspring is produced by mutation.
+    mutpb: Scheduler, default=None
+        An adaptive scheduler representing the probability that an offspring is produced by mutation.
 
     ngen: int, default=None
         The number of generation.
@@ -324,19 +315,11 @@ def eaMuPlusLambda(
         print("INFO: Stopping the algorithm")
         return population, logbook, n_gen
 
-    # Begin the generational process
-    mutpb_value = mutpb
-    cxpb_value = cxpb
     for gen in range(1, ngen + 1):
         try:
-            # Check adaptive rates
-            if isinstance(mutpb, BaseAdapter):
-                mutpb_value = mutpb.step()
-            if isinstance(cxpb, BaseAdapter):
-                cxpb_value = cxpb.step()
 
             # Vary the population
-            offspring = varOr(population, toolbox, lambda_, cxpb_value, mutpb_value)
+            offspring = varOr(population, toolbox, lambda_, cxpb.step(), mutpb.step())
 
             # Evaluate the individuals with an invalid fitness
             invalid_ind = [ind for ind in offspring if not ind.fitness.valid]
@@ -426,11 +409,11 @@ def eaMuCommaLambda(
     lambda\_: int, default=None
         The number of children to produce at each generation.
 
-    cxpb: float or a Scheduler, default=None
+    cxpb: Scheduler, default=None
         The probability that an offspring is produced by crossover.
 
-    mutpb: float or a Scheduler, default=None
-        The probability that an offspring is produced by mutation.
+    mutpb: Scheduler, default=None
+        An adaptive scheduler representing the probability that an offspring is produced by mutation.
 
     ngen: int, default=None
         The number of generation.
@@ -520,19 +503,11 @@ def eaMuCommaLambda(
         print("INFO: Stopping the algorithm")
         return population, logbook, n_gen
 
-    # Begin the generational process
-    mutpb_value = mutpb
-    cxpb_value = cxpb
     for gen in range(1, ngen + 1):
         try:
-            # Check adaptive rates
-            if isinstance(mutpb, BaseAdapter):
-                mutpb_value = mutpb.step()
-            if isinstance(cxpb, BaseAdapter):
-                cxpb_value = cxpb.step()
 
             # Vary the population
-            offspring = varOr(population, toolbox, lambda_, cxpb_value, mutpb_value)
+            offspring = varOr(population, toolbox, lambda_, cxpb.step(), mutpb.step())
 
             # Evaluate the individuals with an invalid fitness
             invalid_ind = [ind for ind in offspring if not ind.fitness.valid]