pydmoo 0.1.1__py3-none-any.whl → 0.1.3__py3-none-any.whl

pydmoo/algorithms/base/core/genetic.py

@@ -92,7 +92,7 @@ class GeneticAlgorithm(Algorithm):
                                         random_state=self.random_state, algorithm=self, **kwargs)
 
     def _infill(self):
-        # Added by DynOpt on Dec 21, 2025
+        # Added by DynOpt Team on Dec 21, 2025
         pop = self._infill_static_dynamic()
 
         # do the mating using the current population
@@ -110,7 +110,7 @@ class GeneticAlgorithm(Algorithm):
 
         return off
 
-    # Added by DynOpt on Dec 21, 2025
+    # Added by DynOpt Team on Dec 21, 2025
     def _infill_static_dynamic(self):
         pop = self.pop
 

pydmoo/algorithms/base/dmoo/dmoead.py

@@ -8,7 +8,21 @@ from pydmoo.algorithms.base.moo.moead import MOEAD
 
 
 class DMOEAD(MOEAD):
-
+    """
+    Dynamic MOEA/D (DMOEAD).
+
+    Extension of MOEAD for dynamic optimization problems.
+
+    Parameters
+    ----------
+    perc_detect_change : float, default=0.1
+        Percentage of population to sample for change detection (0 to 1).
+    eps : float, default=0.0
+        Threshold for change detection. Change is detected when mean squared
+        difference exceeds this value.
+    **kwargs
+        Additional arguments passed to MOEAD parent class.
+    """
     def __init__(self,
                  perc_detect_change=0.1,
                  eps=0.0,
@@ -22,7 +36,15 @@ class DMOEAD(MOEAD):
         assert not problem.has_constraints(), f"{self.__class__.__name__} only works for unconstrained problems."
         return super().setup(problem, **kwargs)
 
-    def _detect_change_sample_part_population(self):
+    def _detect_change_sample_part_population(self) -> bool:
+        """
+        Detect environmental changes by sampling part of the population.
+
+        Returns
+        -------
+        change_detected : bool
+            True if environmental change is detected, False otherwise.
+        """
         pop = self.pop
         X, F = pop.get("X", "F")
 
@@ -40,7 +62,15 @@ class DMOEAD(MOEAD):
         change_detected = delta > self.eps
         return change_detected
 
-    def _next_static_dynamic(self):
+    def _next_static_dynamic(self) -> Population:
+        """
+        Perform next with dynamic change detection and response.
+
+        Returns
+        -------
+        Population
+            Current population after potential response to environmental change.
+        """
         # for dynamic environment
         pop = self.pop
 
@@ -52,7 +82,7 @@ class DMOEAD(MOEAD):
 
                 start_time = time.time()
 
-                pop = self._response_change()
+                pop = self._response_mechanism()
 
                 # reevaluate because we know there was a change
                 self.evaluator.eval(self.problem, pop)
@@ -69,11 +99,25 @@ class DMOEAD(MOEAD):
 
         return pop
 
-    def _response_change(self):
-        pass
+    def _response_mechanism(self):
+        """
+        Response mechanism for environmental change.
+
+        Returns
+        -------
+        Population
+            Population after applying response strategy.
+
+        Raises
+        ------
+        NotImplementedError
+            Must be implemented by subclasses.
+        """
+        raise NotImplementedError
 
 
 class DMOEADA(DMOEAD):
+    """DMOEADA."""
 
     def __init__(self,
                  perc_detect_change=0.1,
@@ -86,7 +130,8 @@ class DMOEADA(DMOEAD):
 
         self.perc_diversity = perc_diversity
 
-    def _response_change(self):
+    def _response_mechanism(self):
+        """Response mechanism."""
         pop = self.pop
         X = pop.get("X")
 
@@ -103,6 +148,7 @@ class DMOEADA(DMOEAD):
 
 
 class DMOEADB(DMOEAD):
+    """DMOEADB."""
 
     def __init__(self,
                  perc_detect_change=0.1,
@@ -115,7 +161,8 @@ class DMOEADB(DMOEAD):
 
         self.perc_diversity = perc_diversity
 
-    def _response_change(self):
+    def _response_mechanism(self):
+        """Response mechanism."""
         pop = self.pop
         X = pop.get("X")
 

pydmoo/algorithms/base/dmoo/dmoeadde.py

@@ -8,6 +8,21 @@ from pydmoo.algorithms.base.moo.moeadde import MOEADDE
 
 
 class DMOEADDE(MOEADDE):
+    """
+    Dynamic MOEA/D-DE (DMOEADDE).
+
+    Extension of MOEADDE for dynamic optimization problems.
+
+    Parameters
+    ----------
+    perc_detect_change : float, default=0.1
+        Percentage of population to sample for change detection (0 to 1).
+    eps : float, default=0.0
+        Threshold for change detection. Change is detected when mean squared
+        difference exceeds this value.
+    **kwargs
+        Additional arguments passed to MOEADDE parent class.
+    """
 
     def __init__(self,
                  perc_detect_change=0.1,
@@ -22,7 +37,15 @@ class DMOEADDE(MOEADDE):
         assert not problem.has_constraints(), f"{self.__class__.__name__} only works for unconstrained problems."
         return super().setup(problem, **kwargs)
 
-    def _detect_change_sample_part_population(self):
+    def _detect_change_sample_part_population(self) -> bool:
+        """
+        Detect environmental changes by sampling part of the population.
+
+        Returns
+        -------
+        change_detected : bool
+            True if environmental change is detected, False otherwise.
+        """
         pop = self.pop
         X, F = pop.get("X", "F")
 
@@ -40,7 +63,15 @@ class DMOEADDE(MOEADDE):
         change_detected = delta > self.eps
         return change_detected
 
-    def _next_static_dynamic(self):
+    def _next_static_dynamic(self) -> Population:
+        """
+        Perform next with dynamic change detection and response.
+
+        Returns
+        -------
+        Population
+            Current population after potential response to environmental change.
+        """
         # for dynamic environment
         pop = self.pop
 
@@ -52,7 +83,7 @@ class DMOEADDE(MOEADDE):
 
                 start_time = time.time()
 
-                pop = self._response_change()
+                pop = self._response_mechanism()
 
                 # reevaluate because we know there was a change
                 self.evaluator.eval(self.problem, pop)
@@ -69,11 +100,25 @@ class DMOEADDE(MOEADDE):
 
         return pop
 
-    def _response_change(self):
-        pass
+    def _response_mechanism(self) -> Population:
+        """
+        Response mechanism for environmental change.
+
+        Returns
+        -------
+        Population
+            Population after applying response strategy.
+
+        Raises
+        ------
+        NotImplementedError
+            Must be implemented by subclasses.
+        """
+        raise NotImplementedError
 
 
 class DMOEADDEA(DMOEADDE):
+    """DMOEADDEA."""
 
     def __init__(self,
                  perc_detect_change=0.1,
@@ -86,7 +131,8 @@ class DMOEADDEA(DMOEADDE):
 
         self.perc_diversity = perc_diversity
 
-    def _response_change(self):
+    def _response_mechanism(self):
+        """Response mechanism."""
         pop = self.pop
         X = pop.get("X")
 
@@ -103,6 +149,7 @@ class DMOEADDEA(DMOEADDE):
 
 
 class DMOEADDEB(DMOEADDE):
+    """DMOEADDEB."""
 
     def __init__(self,
                  perc_detect_change=0.1,
@@ -115,7 +162,8 @@ class DMOEADDEB(DMOEADDE):
 
         self.perc_diversity = perc_diversity
 
-    def _response_change(self):
+    def _response_mechanism(self):
+        """Response mechanism."""
         pop = self.pop
         X = pop.get("X")
 

pydmoo/algorithms/base/dmoo/dnsga2.py

@@ -17,10 +17,25 @@ from pydmoo.algorithms.base.moo.nsga2 import NSGA2
 
 
 class DNSGA2(NSGA2):
+    """
+    Dynamic Non-dominated Sorting Genetic Algorithm II (DNSGA2).
+
+    Extension of NSGA2 for dynamic optimization problems.
+
+    Parameters
+    ----------
+    perc_detect_change : float, default=0.1
+        Percentage of population to sample for change detection (0 to 1).
+    eps : float, default=0.0
+        Threshold for change detection. Change is detected when mean squared
+        difference exceeds this value.
+    **kwargs
+        Additional arguments passed to NSGA2 parent class.
+    """
 
     def __init__(self,
-                 perc_detect_change=0.1,
-                 eps=0.0,
+                 perc_detect_change: float = 0.1,
+                 eps: float = 0.0,
                  **kwargs):
 
         super().__init__(**kwargs)
@@ -31,7 +46,15 @@ class DNSGA2(NSGA2):
         assert not problem.has_constraints(), f"{self.__class__.__name__} only works for unconstrained problems."
         return super().setup(problem, **kwargs)
 
-    def _detect_change_sample_part_population(self):
+    def _detect_change_sample_part_population(self) -> bool:
+        """
+        Detect environmental changes by sampling part of the population.
+
+        Returns
+        -------
+        change_detected : bool
+            True if environmental change is detected, False otherwise.
+        """
         pop = self.pop
         X, F = pop.get("X", "F")
 
@@ -49,7 +72,15 @@ class DNSGA2(NSGA2):
         change_detected = delta > self.eps
         return change_detected
 
-    def _infill_static_dynamic(self):
+    def _infill_static_dynamic(self) -> Population:
+        """
+        Perform infill with dynamic change detection and response.
+
+        Returns
+        -------
+        Population
+            Current population after potential response to environmental change.
+        """
         # for dynamic environment
         pop = self.pop
 
@@ -59,7 +90,7 @@ class DNSGA2(NSGA2):
 
             start_time = time.time()
 
-            pop = self._response_change()
+            pop = self._response_mechanism()
 
             # reevaluate because we know there was a change
             self.evaluator.eval(self.problem, pop)
@@ -75,16 +106,39 @@ class DNSGA2(NSGA2):
 
         return pop
 
-    def _response_change(self):
-        pass
+    def _response_mechanism(self) -> Population:
+        """
+        Response mechanism for environmental change.
+
+        Returns
+        -------
+        Population
+            Population after applying response strategy.
+
+        Raises
+        ------
+        NotImplementedError
+            Must be implemented by subclasses.
+        """
+        raise NotImplementedError
 
 
 class DNSGA2A(DNSGA2):
+    """DNSGA2A.
+
+    References
+    ----------
+    Deb, K., Rao N., U. B., and Karthik, S. (2007).
+    Dynamic multi-objective optimization and decision-making using modified NSGA-II:
+        A case study on hydro-thermal power scheduling.
+    Evolutionary Multi-Criterion Optimization, 803–817.
+    https://doi.org/10.1007/978-3-540-70928-2_60
+    """
 
     def __init__(self,
-                 perc_detect_change=0.1,
-                 eps=0.0,
-                 perc_diversity=0.3,
+                 perc_detect_change: float = 0.1,
+                 eps: float = 0.0,
+                 perc_diversity: float = 0.3,
                  **kwargs):
         super().__init__(perc_detect_change=perc_detect_change,
                          eps=eps,
@@ -92,7 +146,8 @@ class DNSGA2A(DNSGA2):
 
         self.perc_diversity = perc_diversity
 
-    def _response_change(self):
+    def _response_mechanism(self) -> Population:
+        """Response mechanism."""
         pop = self.pop
         X = pop.get("X")
 
@@ -109,11 +164,21 @@ class DNSGA2A(DNSGA2):
 
 
 class DNSGA2B(DNSGA2):
+    """DNSGA2B.
+
+    References
+    ----------
+    Deb, K., Rao N., U. B., and Karthik, S. (2007).
+    Dynamic multi-objective optimization and decision-making using modified NSGA-II:
+        A case study on hydro-thermal power scheduling.
+    Evolutionary Multi-Criterion Optimization, 803–817.
+    https://doi.org/10.1007/978-3-540-70928-2_60
+    """
 
     def __init__(self,
-                 perc_detect_change=0.1,
-                 eps=0.0,
-                 perc_diversity=0.3,
+                 perc_detect_change: float = 0.1,
+                 eps: float = 0.0,
+                 perc_diversity: float = 0.3,
                  **kwargs):
         super().__init__(perc_detect_change=perc_detect_change,
                          eps=eps,
@@ -121,7 +186,8 @@ class DNSGA2B(DNSGA2):
 
         self.perc_diversity = perc_diversity
 
-    def _response_change(self):
+    def _response_mechanism(self) -> Population:
+        """Response mechanism."""
         pop = self.pop
         X = pop.get("X")
 

pydmoo/algorithms/base/moo/moead.py

@@ -105,6 +105,7 @@ class MOEAD(LoopwiseAlgorithm, GeneticAlgorithm):
         self.ideal = np.min(self.pop.get("F"), axis=0)
 
     def _next(self):
+        # Added by DynOpt Team on Dec 21, 2025
         pop = self._next_static_dynamic()
 
         # iterate for each member of the population in random order
@@ -124,7 +125,7 @@ class MOEAD(LoopwiseAlgorithm, GeneticAlgorithm):
             # now actually do the replacement of the individual is better
             self._replace(k, off)
 
-    # Added by DynOpt on Dec 21, 2025
+    # Added by DynOpt Team on Dec 21, 2025
     def _next_static_dynamic(self):
         pop = self.pop
 

pydmoo/algorithms/base/moo/moeadde.py

@@ -7,13 +7,18 @@ from pydmoo.algorithms.base.moo.moead import MOEAD
 
 
 class MOEADDE(MOEAD):
-    """MOEA/D-DE (Updated by Cao).
-
-    It is worth noting that there is a distinct modification in line 28 compared with the original framework of MOEA/D-DE.
-    The newly generated solution competes with each member from the corresponding mating neighborhood (denoted as Pool in Algorithm 2).
-    But in the original MOEA/D-DE framework, it only competes with two members from the corresponding mating neighborhood.
-    This modification expands the replacement neighborhood to enhance the exploitation capability that is extremely important in dealing with DMOPs.
-
+    """MOEA/D-DE.
+
+    Notes
+    -----
+    It is worth noting that there is a distinct modification in line 28 compared with the original framework of
+    MOEA/D-DE. The newly generated solution competes with each member from the corresponding mating neighborhood
+    (denoted as Pool in Algorithm 2). But in the original MOEA/D-DE framework, it only competes with two members from
+    the corresponding mating neighborhood. This modification expands the replacement neighborhood to enhance the
+    exploitation capability that is extremely important in dealing with DMOPs (Cao et al., 2020).
+
+    References
+    ----------
     Cao, L., Xu, L., Goodman, E. D., Bao, C., and Zhu, S. (2020).
     Evolutionary dynamic multiobjective optimization assisted by a support vector regression predictor.
     IEEE Transactions on Evolutionary Computation, 24(2), 305–319.

pydmoo/algorithms/classic/moead_ae.py

@@ -21,8 +21,8 @@ class MOEADAE(DMOEAD):
 
         super().__init__(**kwargs)
 
-    def _response_change(self):
-        """Response."""
+    def _response_mechanism(self):
+        """Response mechanism."""
         pop = self.pop
         X = pop.get("X")
 

pydmoo/algorithms/classic/moead_pps.py

@@ -25,8 +25,8 @@ class MOEADPPS(DMOEAD):
         self.p = 3  # the order of the AR model
         self.M = 23  # the length of history mean point series
 
-    def _response_change(self):
-        """Response."""
+    def _response_mechanism(self):
+        """Response mechanism."""
         pop = self.pop
         X = pop.get("X")
 

pydmoo/algorithms/classic/moead_svr.py

@@ -0,0 +1,86 @@
+import numpy as np
+from pymoo.core.population import Population
+from sklearn.svm import SVR
+
+from pydmoo.algorithms.base.dmoo.dmoead import DMOEAD
+
+
+class MOEADSVR(DMOEAD):
+    """Support Vector Regression (SVR).
+
+    Notes
+    -----
+    [Official Python Code](https://github.com/LeileiCao/MOEA-D-SVR/blob/master/MOEAD-SVR%20.py)
+
+    References
+    ----------
+    Cao, L., Xu, L., Goodman, E. D., Bao, C., and Zhu, S. (2020).
+    Evolutionary dynamic multiobjective optimization assisted by a support vector regression predictor.
+    IEEE Transactions on Evolutionary Computation, 24(2), 305–319.
+    https://doi.org/10.1109/TEVC.2019.2925722
+    """
+
+    def __init__(self, **kwargs):
+        super().__init__(**kwargs)
+
+        # SVR
+        self._q = 4  # the number of preceding values that are correlated with the target value (dimension of input samples in the SVR model)
+        self._C = 1000  # the regularization constant in SVR model
+        self._epsilon = 0.05  # the insensitive tube size in SVR model
+        # self._gamma = 1/d  # the Gaussian RBF kernel parameter used in SVR model, and d is the number of variables
+
+    def _response_mechanism(self):
+        """Response mechanism."""
+        pop = self.pop
+        X = pop.get("X")
+
+        old = self.data.get("stacked_X", None)
+        if old is None:
+            stacked_X = np.expand_dims(X, axis=0)
+        else:
+            stacked_X = np.concatenate((old, np.expand_dims(X, axis=0)), axis=0)
+        self.data["stacked_X"] = stacked_X
+
+        N, d = X.shape
+        sol = np.zeros((N, d))
+        K = len(stacked_X)
+
+        if K < self._q + 2:
+            # recreate the current population without being evaluated
+            # Re-evaluate the current population, and update the reference point
+            pop = Population.new(X=X)
+
+            return pop
+
+        # Precompute sliding window indices to avoid redundant calculations
+        window_indices = np.lib.stride_tricks.sliding_window_view(np.arange(K), self._q + 1)
+
+        for i in range(N):
+            for j in range(d):
+                # Extract the time series for this (i,j) position
+                ts = stacked_X[:K, i, j]
+
+                # Create training data using vectorized sliding windows
+                train = ts[window_indices]
+                x_train = train[:, :-1]
+                y_train = train[:, -1]
+
+                # Train SVR model (consider moving this outside loops if possible)
+                # gamma if 'auto', uses 1 / n_features (not provided in code but provided in paper)
+                # versionchanged:: 0.22
+                # The default value of ``gamma`` changed from 'auto' to 'scale'.
+                svr = SVR(kernel='rbf', epsilon=self._epsilon, C=self._C, gamma=1/d)
+                model = svr.fit(x_train, y_train)
+
+                # Make prediction
+                sol[i, j] = model.predict(ts[-self._q:].reshape(1, -1))
+
+        # bounds
+        if self.problem.has_bounds():
+            xl, xu = self.problem.bounds()
+            sol = np.clip(sol, xl, xu)  # provided in the original reference literature
+
+        # recreate the current population without being evaluated
+        pop = Population.new(X=sol)
+
+        return pop

pydmoo/algorithms/classic/moeadde_ae.py

@@ -21,8 +21,8 @@ class MOEADDEAE(DMOEADDE):
 
         super().__init__(**kwargs)
 
-    def _response_change(self):
-        """Response."""
+    def _response_mechanism(self):
+        """Response mechanism."""
         pop = self.pop
         X = pop.get("X")
 

pydmoo/algorithms/classic/moeadde_pps.py

@@ -25,8 +25,8 @@ class MOEADDEPPS(DMOEADDE):
         self.p = 3  # the order of the AR model
         self.M = 23  # the length of history mean point series
 
-    def _response_change(self):
-        """Response."""
+    def _response_mechanism(self):
+        """Response mechanism."""
         pop = self.pop
         X = pop.get("X")
 

pydmoo/algorithms/classic/moeadde_svr.py

@@ -0,0 +1,86 @@
+import numpy as np
+from pymoo.core.population import Population
+from sklearn.svm import SVR
+
+from pydmoo.algorithms.base.dmoo.dmoeadde import DMOEADDE
+
+
+class MOEADDESVR(DMOEADDE):
+    """Support Vector Regression (SVR).
+
+    Notes
+    -----
+    [Official Python Code](https://github.com/LeileiCao/MOEA-D-SVR/blob/master/MOEAD-SVR%20.py)
+
+    References
+    ----------
+    Cao, L., Xu, L., Goodman, E. D., Bao, C., and Zhu, S. (2020).
+    Evolutionary dynamic multiobjective optimization assisted by a support vector regression predictor.
+    IEEE Transactions on Evolutionary Computation, 24(2), 305–319.
+    https://doi.org/10.1109/TEVC.2019.2925722
+    """
+
+    def __init__(self, **kwargs):
+        super().__init__(**kwargs)
+
+        # SVR
+        self._q = 4  # the number of preceding values that are correlated with the target value (dimension of input samples in the SVR model)
+        self._C = 1000  # the regularization constant in SVR model
+        self._epsilon = 0.05  # the insensitive tube size in SVR model
+        # self._gamma = 1/d  # the Gaussian RBF kernel parameter used in SVR model, and d is the number of variables
+
+    def _response_mechanism(self):
+        """Response mechanism."""
+        pop = self.pop
+        X = pop.get("X")
+
+        old = self.data.get("stacked_X", None)
+        if old is None:
+            stacked_X = np.expand_dims(X, axis=0)
+        else:
+            stacked_X = np.concatenate((old, np.expand_dims(X, axis=0)), axis=0)
+        self.data["stacked_X"] = stacked_X
+
+        N, d = X.shape
+        sol = np.zeros((N, d))
+        K = len(stacked_X)
+
+        if K < self._q + 2:
+            # recreate the current population without being evaluated
+            # Re-evaluate the current population, and update the reference point
+            pop = Population.new(X=X)
+
+            return pop
+
+        # Precompute sliding window indices to avoid redundant calculations
+        window_indices = np.lib.stride_tricks.sliding_window_view(np.arange(K), self._q + 1)
+
+        for i in range(N):
+            for j in range(d):
+                # Extract the time series for this (i,j) position
+                ts = stacked_X[:K, i, j]
+
+                # Create training data using vectorized sliding windows
+                train = ts[window_indices]
+                x_train = train[:, :-1]
+                y_train = train[:, -1]
+
+                # Train SVR model (consider moving this outside loops if possible)
+                # gamma if 'auto', uses 1 / n_features (not provided in code but provided in paper)
+                # versionchanged:: 0.22
+                # The default value of ``gamma`` changed from 'auto' to 'scale'.
+                svr = SVR(kernel='rbf', epsilon=self._epsilon, C=self._C, gamma=1/d)
+                model = svr.fit(x_train, y_train)
+
+                # Make prediction
+                sol[i, j] = model.predict(ts[-self._q:].reshape(1, -1))[0]
+
+        # bounds
+        if self.problem.has_bounds():
+            xl, xu = self.problem.bounds()
+            sol = np.clip(sol, xl, xu)  # provided in the original reference literature
+
+        # recreate the current population without being evaluated
+        pop = Population.new(X=sol)
+
+        return pop