shapiq 1.2.1__tar.gz → 1.2.2__tar.gz

{shapiq-1.2.1 → shapiq-1.2.2}/CHANGELOG.md

@@ -1,7 +1,11 @@
 ## Changelog
 
 ### Development
-...
+
+### v1.2.2 (2025-03-11)
+- changes python support to 3.10-3.13 [#318](https://github.com/mmschlk/shapiq/pull/318)
+- fixes a bug that prohibited importing shapiq in environments without write access [#326](https://github.com/mmschlk/shapiq/issues/326)
+- adds `ExtraTreeRegressors` to supported models [#309](https://github.com/mmschlk/shapiq/pull/309)
 
 ### v1.2.1 (2025-02-17)
 - fixes bugs regarding plotting [#315](https://github.com/mmschlk/shapiq/issues/315) and [#316](https://github.com/mmschlk/shapiq/issues/316)

{shapiq-1.2.1/shapiq.egg-info → shapiq-1.2.2}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.2
 Name: shapiq
-Version: 1.2.1
+Version: 1.2.2
 Summary: Shapley Interactions for Machine Learning
 Home-page: https://github.com/mmschlk/shapiq
 Author: Maximilian Muschalik et al.
@@ -17,11 +17,11 @@ Classifier: Intended Audience :: Science/Research
 Classifier: Topic :: Scientific/Engineering
 Classifier: Topic :: Scientific/Engineering :: Artificial Intelligence
 Classifier: Programming Language :: Python
-Classifier: Programming Language :: Python :: 3.9
 Classifier: Programming Language :: Python :: 3.10
 Classifier: Programming Language :: Python :: 3.11
 Classifier: Programming Language :: Python :: 3.12
-Requires-Python: >=3.9.0
+Classifier: Programming Language :: Python :: 3.13
+Requires-Python: >=3.10.0
 Description-Content-Type: text/markdown
 License-File: LICENSE
 Requires-Dist: numpy
@@ -108,7 +108,7 @@ Dynamic: summary
 Shapley Interaction Quantification (`shapiq`) is a Python package for (1) approximating any-order Shapley interactions, (2) benchmarking game-theoretical algorithms for machine learning, (3) explaining feature interactions of model predictions. `shapiq` extends the well-known [shap](https://github.com/shap/shap) package for both researchers working on game theory in machine learning, as well as the end-users explaining models. SHAP-IQ extends individual Shapley values by quantifying the **synergy** effect between entities (aka **players** in the jargon of game theory) like explanatory features, data points, or weak learners in ensemble models. Synergies between players give a more comprehensive view of machine learning models.
 
 ## 🛠️ Install
-`shapiq` is intended to work with **Python 3.9 and above**. Installation can be done via `pip`:
+`shapiq` is intended to work with **Python 3.10 and above**. Installation can be done via `pip`:
 
 ```sh
 pip install shapiq
@@ -283,7 +283,11 @@ Built with ❤️ by the shapiq team.
 ## Changelog
 
 ### Development
-...
+
+### v1.2.2 (2025-03-11)
+- changes python support to 3.10-3.13 [#318](https://github.com/mmschlk/shapiq/pull/318)
+- fixes a bug that prohibited importing shapiq in environments without write access [#326](https://github.com/mmschlk/shapiq/issues/326)
+- adds `ExtraTreeRegressors` to supported models [#309](https://github.com/mmschlk/shapiq/pull/309)
 
 ### v1.2.1 (2025-02-17)
 - fixes bugs regarding plotting [#315](https://github.com/mmschlk/shapiq/issues/315) and [#316](https://github.com/mmschlk/shapiq/issues/316)

{shapiq-1.2.1 → shapiq-1.2.2}/README.md

@@ -17,7 +17,7 @@
 Shapley Interaction Quantification (`shapiq`) is a Python package for (1) approximating any-order Shapley interactions, (2) benchmarking game-theoretical algorithms for machine learning, (3) explaining feature interactions of model predictions. `shapiq` extends the well-known [shap](https://github.com/shap/shap) package for both researchers working on game theory in machine learning, as well as the end-users explaining models. SHAP-IQ extends individual Shapley values by quantifying the **synergy** effect between entities (aka **players** in the jargon of game theory) like explanatory features, data points, or weak learners in ensemble models. Synergies between players give a more comprehensive view of machine learning models.
 
 ## 🛠️ Install
-`shapiq` is intended to work with **Python 3.9 and above**. Installation can be done via `pip`:
+`shapiq` is intended to work with **Python 3.10 and above**. Installation can be done via `pip`:
 
 ```sh
 pip install shapiq

{shapiq-1.2.1 → shapiq-1.2.2}/pyproject.toml

@@ -7,10 +7,10 @@ testpaths = "tests"
 
 [tool.black]
 line-length = 100
-target-version = ['py39', 'py310', 'py311']
+target-version = ['py310', 'py311', 'py312', 'py313']
 
 [tool.ruff]
 lint.select = ["E", "F", "I", "UP"]  # https://beta.ruff.rs/docs/rules/
 lint.ignore = ["E501"]
 line-length = 100
-target-version = "py39"
+target-version = "py310"

{shapiq-1.2.1 → shapiq-1.2.2}/setup.py

@@ -9,7 +9,7 @@ LONG_DESCRIPTION_CONTENT_TYPE = "text/markdown"
 URL = "https://github.com/mmschlk/shapiq"
 EMAIL = "maximilian.muschalik@ifi.lmu.de"
 AUTHOR = "Maximilian Muschalik et al."
-REQUIRES_PYTHON = ">=3.9.0"
+REQUIRES_PYTHON = ">=3.10.0"
 
 work_directory = os.path.abspath(os.path.dirname(__file__))
 
@@ -89,10 +89,10 @@ setuptools.setup(
         "Topic :: Scientific/Engineering",
         "Topic :: Scientific/Engineering :: Artificial Intelligence",
         "Programming Language :: Python",
-        "Programming Language :: Python :: 3.9",
         "Programming Language :: Python :: 3.10",
         "Programming Language :: Python :: 3.11",
         "Programming Language :: Python :: 3.12",
+        "Programming Language :: Python :: 3.13",
     ],
     keywords=[
         "python",

{shapiq-1.2.1 → shapiq-1.2.2}/shapiq/__init__.py

@@ -2,7 +2,7 @@
 the well established Shapley value and its generalization to interaction.
 """
 
-__version__ = "1.2.1"
+__version__ = "1.2.2"
 
 # approximator classes
 from .approximator import (

{shapiq-1.2.1 → shapiq-1.2.2}/shapiq/approximator/_base.py

@@ -2,7 +2,7 @@
 
 import copy
 from abc import ABC, abstractmethod
-from typing import Callable, Optional
+from collections.abc import Callable
 
 import numpy as np
 
@@ -66,8 +66,8 @@ class Approximator(ABC):
         top_order: bool,
         min_order: int = 0,
         pairing_trick: bool = False,
-        sampling_weights: Optional[np.ndarray[float]] = None,
-        random_state: Optional[int] = None,
+        sampling_weights: np.ndarray[float] | None = None,
+        random_state: int | None = None,
     ) -> None:
         # check if index can be approximated
         self.index: str = index
@@ -92,8 +92,8 @@ class Approximator(ABC):
         )
 
         # set up random state and random number generators
-        self._random_state: Optional[int] = random_state
-        self._rng: Optional[np.random.Generator] = np.random.default_rng(seed=self._random_state)
+        self._random_state: int | None = random_state
+        self._rng: np.random.Generator | None = np.random.default_rng(seed=self._random_state)
 
         # set up a coalition sampler
         self._big_M: int = 100_000_000  # large number for sampling weights
@@ -178,8 +178,8 @@ class Approximator(ABC):
         result,
         baseline_value: float,
         *,
-        estimated: Optional[bool] = None,
-        budget: Optional[int] = None,
+        estimated: bool | None = None,
+        budget: int | None = None,
     ) -> InteractionValues:
         """Finalizes the result dictionary.
 
@@ -303,8 +303,8 @@ class Approximator(ABC):
     @staticmethod
     def aggregate_interaction_values(
         base_interactions: InteractionValues,
-        order: Optional[int] = None,
-        player_set: Optional[set[int]] = None,
+        order: int | None = None,
+        player_set: set[int] | None = None,
     ) -> InteractionValues:
         """Aggregates the interaction values.
 

{shapiq-1.2.1 → shapiq-1.2.2}/shapiq/approximator/marginals/owen.py

@@ -2,7 +2,7 @@
 Okhrati and Lipani (2020). It estimates the Shapley values in its integral representation by
 sampling random marginal contributions."""
 
-from typing import Callable, Optional
+from collections.abc import Callable
 
 import numpy as np
 
@@ -34,7 +34,7 @@ class OwenSamplingSV(Approximator):
         self,
         n: int,
         n_anchor_points: int = 10,
-        random_state: Optional[int] = None,
+        random_state: int | None = None,
     ) -> None:
         super().__init__(n, max_order=1, index="SV", top_order=False, random_state=random_state)
         self.iteration_cost: int = 2

{shapiq-1.2.1 → shapiq-1.2.2}/shapiq/approximator/marginals/stratified.py

@@ -1,6 +1,6 @@
 """This module contains the Stratified Sampling approximation method for the Shapley values."""
 
-from typing import Callable, Optional
+from collections.abc import Callable
 
 import numpy as np
 
@@ -31,7 +31,7 @@ class StratifiedSamplingSV(Approximator):
     def __init__(
         self,
         n: int,
-        random_state: Optional[int] = None,
+        random_state: int | None = None,
     ) -> None:
         super().__init__(n, max_order=1, index="SV", top_order=False, random_state=random_state)
         self.iteration_cost: int = 2

{shapiq-1.2.1 → shapiq-1.2.2}/shapiq/approximator/montecarlo/_base.py

@@ -1,6 +1,6 @@
 """This module contains the Base Regression approximator to compute SII and k-SII of arbitrary max_order."""
 
-from typing import Callable, Optional
+from collections.abc import Callable
 
 import numpy as np
 from scipy.special import binom, factorial
@@ -41,7 +41,7 @@ class MonteCarlo(Approximator):
         stratify_coalition_size: bool = True,
         stratify_intersection: bool = True,
         top_order: bool = False,
-        random_state: Optional[int] = None,
+        random_state: int | None = None,
         pairing_trick: bool = False,
         sampling_weights: np.ndarray = None,
     ):

{shapiq-1.2.1 → shapiq-1.2.2}/shapiq/approximator/montecarlo/shapiq.py

@@ -2,8 +2,6 @@
 The Unbiased KernelSHAP method is a variant of KernelSHAP. However, it was shown that Unbiased
 KernelSHAP is a more specific variant of the ShapIQ interaction method."""
 
-from typing import Optional
-
 from ._base import MonteCarlo
 
 
@@ -53,9 +51,9 @@ class SHAPIQ(MonteCarlo):
         max_order: int = 2,
         index: str = "k-SII",
         top_order: bool = False,
-        sampling_weights: Optional[float] = None,
+        sampling_weights: float | None = None,
         pairing_trick: bool = False,
-        random_state: Optional[int] = None,
+        random_state: int | None = None,
     ):
         super().__init__(
             n,
@@ -117,8 +115,8 @@ class UnbiasedKernelSHAP(SHAPIQ):
         self,
         n: int,
         pairing_trick: bool = False,
-        sampling_weights: Optional[float] = None,
-        random_state: Optional[int] = None,
+        sampling_weights: float | None = None,
+        random_state: int | None = None,
         **kwargs,
     ):
         super().__init__(

{shapiq-1.2.1 → shapiq-1.2.2}/shapiq/approximator/montecarlo/svarmiq.py

@@ -1,7 +1,5 @@
 """SVARM-IQ approximation."""
 
-from typing import Optional
-
 from ._base import MonteCarlo
 
 
@@ -37,8 +35,8 @@ class SVARMIQ(MonteCarlo):
         index: str = "k-SII",
         top_order: bool = False,
         pairing_trick: bool = False,
-        sampling_weights: Optional[float] = None,
-        random_state: Optional[int] = None,
+        sampling_weights: float | None = None,
+        random_state: int | None = None,
     ):
         super().__init__(
             n,
@@ -75,9 +73,9 @@ class SVARM(SVARMIQ):
     def __init__(
         self,
         n: int,
-        random_state: Optional[int] = None,
+        random_state: int | None = None,
         pairing_trick: bool = False,
-        sampling_weights: Optional[float] = None,
+        sampling_weights: float | None = None,
         **kwargs,
     ):
         super().__init__(

{shapiq-1.2.1 → shapiq-1.2.2}/shapiq/approximator/permutation/sii.py

@@ -1,6 +1,6 @@
 """This module implements the Permutation Sampling approximator for the SII (and k-SII) index."""
 
-from typing import Callable, Optional
+from collections.abc import Callable
 
 import numpy as np
 
@@ -33,7 +33,7 @@ class PermutationSamplingSII(Approximator):
         max_order: int = 2,
         index: str = "SII",
         top_order: bool = False,
-        random_state: Optional[int] = None,
+        random_state: int | None = None,
     ) -> None:
         if index not in ["SII", "k-SII"]:
             raise ValueError(f"Invalid index {index}. Must be either 'SII' or 'k-SII'.")
@@ -68,7 +68,7 @@ class PermutationSamplingSII(Approximator):
         self,
         budget: int,
         game: Callable[[np.ndarray], np.ndarray],
-        batch_size: Optional[int] = 5,
+        batch_size: int | None = 5,
     ) -> InteractionValues:
         """Approximates the interaction values.
 

{shapiq-1.2.1 → shapiq-1.2.2}/shapiq/approximator/permutation/stii.py

@@ -1,7 +1,7 @@
 """This module contains the permutation sampling algorithms to estimate STII scores."""
 
 import warnings
-from typing import Callable, Optional
+from collections.abc import Callable
 
 import numpy as np
 import scipy as sp
@@ -54,7 +54,7 @@ class PermutationSamplingSTII(Approximator):
         )
     """
 
-    def __init__(self, n: int, max_order: int, random_state: Optional[int] = None) -> None:
+    def __init__(self, n: int, max_order: int, random_state: int | None = None) -> None:
         super().__init__(
             n=n, max_order=max_order, index="STII", top_order=False, random_state=random_state
         )

{shapiq-1.2.1 → shapiq-1.2.2}/shapiq/approximator/permutation/sv.py

@@ -1,6 +1,6 @@
 """This module contains the permutation sampling approximation method for the Shapley value (SV)."""
 
-from typing import Callable, Optional
+from collections.abc import Callable
 
 import numpy as np
 
@@ -30,12 +30,12 @@ class PermutationSamplingSV(Approximator):
 
     """
 
-    def __init__(self, n: int, random_state: Optional[int] = None, **kwargs) -> None:
+    def __init__(self, n: int, random_state: int | None = None, **kwargs) -> None:
         super().__init__(n=n, max_order=1, index="SV", top_order=False, random_state=random_state)
         self.iteration_cost: int = n - 1
 
     def approximate(
-        self, budget: int, game: Callable[[np.ndarray], np.ndarray], batch_size: Optional[int] = 5
+        self, budget: int, game: Callable[[np.ndarray], np.ndarray], batch_size: int | None = 5
     ) -> InteractionValues:
         """Approximates the Shapley values using ApproShapley.
 

{shapiq-1.2.1 → shapiq-1.2.2}/shapiq/approximator/regression/_base.py

@@ -2,7 +2,7 @@
 
 import copy
 import warnings
-from typing import Callable, Optional
+from collections.abc import Callable
 
 import numpy as np
 from scipy.special import bernoulli, binom
@@ -42,8 +42,8 @@ class Regression(Approximator):
         index: str,
         sii_consistent: bool = True,
         pairing_trick: bool = False,
-        sampling_weights: Optional[np.ndarray] = None,
-        random_state: Optional[int] = None,
+        sampling_weights: np.ndarray | None = None,
+        random_state: int | None = None,
     ):
         if index not in AVAILABLE_INDICES_REGRESSION:
             raise ValueError(

{shapiq-1.2.1 → shapiq-1.2.2}/shapiq/approximator/regression/fsi.py

@@ -1,7 +1,5 @@
 """Regression with Shapley interaction index (SII) approximation."""
 
-from typing import Optional
-
 import numpy as np
 
 from ._base import Regression
@@ -40,8 +38,8 @@ class RegressionFSII(Regression):
         n: int,
         max_order: int = 2,
         pairing_trick: bool = False,
-        sampling_weights: Optional[np.ndarray] = None,
-        random_state: Optional[int] = None,
+        sampling_weights: np.ndarray | None = None,
+        random_state: int | None = None,
     ):
         super().__init__(
             n,

{shapiq-1.2.1 → shapiq-1.2.2}/shapiq/approximator/regression/kadd_shap.py

@@ -1,7 +1,5 @@
 """Regression with Shapley interaction index (SII) approximation."""
 
-from typing import Optional
-
 import numpy as np
 
 from ._base import Regression
@@ -43,8 +41,8 @@ class kADDSHAP(Regression):
         n: int,
         max_order: int = 2,
         pairing_trick: bool = False,
-        sampling_weights: Optional[np.ndarray] = None,
-        random_state: Optional[int] = None,
+        sampling_weights: np.ndarray | None = None,
+        random_state: int | None = None,
     ):
         super().__init__(
             n,

{shapiq-1.2.1 → shapiq-1.2.2}/shapiq/approximator/regression/kernelshap.py

@@ -1,7 +1,5 @@
 """This module contains the KernelSHAP regression approximator for estimating the SV."""
 
-from typing import Optional
-
 import numpy as np
 
 from ._base import Regression
@@ -55,8 +53,8 @@ class KernelSHAP(Regression):
         self,
         n: int,
         pairing_trick: bool = False,
-        sampling_weights: Optional[np.ndarray] = None,
-        random_state: Optional[int] = None,
+        sampling_weights: np.ndarray | None = None,
+        random_state: int | None = None,
         **kwargs,
     ):
         super().__init__(

{shapiq-1.2.1 → shapiq-1.2.2}/shapiq/approximator/regression/kernelshapiq.py

@@ -1,7 +1,5 @@
 """Regression with Shapley interaction index (SII) approximation."""
 
-from typing import Optional
-
 import numpy as np
 
 from ._base import Regression
@@ -55,8 +53,8 @@ class KernelSHAPIQ(Regression):
         max_order: int = 2,
         index: str = "k-SII",
         pairing_trick: bool = False,
-        sampling_weights: Optional[np.ndarray] = None,
-        random_state: Optional[int] = None,
+        sampling_weights: np.ndarray | None = None,
+        random_state: int | None = None,
     ) -> None:
         if index not in AVAILABLE_INDICES_KERNELSHAPIQ:
             raise ValueError(
@@ -115,8 +113,8 @@ class InconsistentKernelSHAPIQ(Regression):
         max_order: int = 2,
         index: str = "k-SII",
         pairing_trick: bool = False,
-        sampling_weights: Optional[np.ndarray] = None,
-        random_state: Optional[int] = None,
+        sampling_weights: np.ndarray | None = None,
+        random_state: int | None = None,
     ) -> None:
         if index not in AVAILABLE_INDICES_KERNELSHAPIQ:
             raise ValueError(

{shapiq-1.2.1 → shapiq-1.2.2}/shapiq/approximator/sampling.py

@@ -2,7 +2,6 @@
 
 import copy
 import warnings
-from typing import Optional
 
 import numpy as np
 from scipy.special import binom
@@ -76,7 +75,7 @@ class CoalitionSampler:
         n_players: int,
         sampling_weights: np.ndarray,
         pairing_trick: bool = False,
-        random_state: Optional[int] = None,
+        random_state: int | None = None,
     ) -> None:
         self.pairing_trick: bool = pairing_trick
 
@@ -115,28 +114,26 @@ class CoalitionSampler:
             if weight == 0 and 0 < size < self.n:
                 self.n_max_coalitions -= int(binom(self.n, size))
                 self._coalitions_to_exclude.extend([size])
-        self.adjusted_sampling_weights: Optional[np.ndarray[float]] = None
+        self.adjusted_sampling_weights: np.ndarray[float] | None = None
 
         # set sample size variables (for border trick)
-        self._coalitions_to_compute: Optional[list[int]] = None  # coalitions to compute
-        self._coalitions_to_sample: Optional[list[int]] = None  # coalitions to sample
+        self._coalitions_to_compute: list[int] | None = None  # coalitions to compute
+        self._coalitions_to_sample: list[int] | None = None  # coalitions to sample
 
         # initialize variables to be computed and stored
-        self.sampled_coalitions_dict: Optional[dict[tuple[int, ...], int]] = None  # coal -> count
-        self.coalitions_per_size: Optional[np.ndarray[int]] = None  # number of coalitions per size
+        self.sampled_coalitions_dict: dict[tuple[int, ...], int] | None = None  # coal -> count
+        self.coalitions_per_size: np.ndarray[int] | None = None  # number of coalitions per size
 
         # variables accessible through properties
-        self._sampled_coalitions_matrix: Optional[np.ndarray[bool]] = None  # coalitions
-        self._sampled_coalitions_counter: Optional[np.ndarray[int]] = None  # coalitions_counter
-        self._sampled_coalitions_size_prob: Optional[np.ndarray[float]] = (
+        self._sampled_coalitions_matrix: np.ndarray[bool] | None = None  # coalitions
+        self._sampled_coalitions_counter: np.ndarray[int] | None = None  # coalitions_counter
+        self._sampled_coalitions_size_prob: np.ndarray[float] | None = (
             None  # coalitions_size_probability
         )
-        self._sampled_coalitions_in_size_prob: Optional[np.ndarray[float]] = (
+        self._sampled_coalitions_in_size_prob: np.ndarray[float] | None = (
             None  # coalitions_in_size_probability
         )
-        self._is_coalition_size_sampled: Optional[np.ndarray[bool]] = (
-            None  # is_coalition_size_sampled
-        )
+        self._is_coalition_size_sampled: np.ndarray[bool] | None = None  # is_coalition_size_sampled
 
     @property
     def n_coalitions(self) -> int:
@@ -272,7 +269,7 @@ class CoalitionSampler:
         return np.sum(self.coalitions_matrix, axis=1)
 
     @property
-    def empty_coalition_index(self) -> Optional[int]:
+    def empty_coalition_index(self) -> int | None:
         """Returns the index of the empty coalition.
 
         Returns:

{shapiq-1.2.1 → shapiq-1.2.2}/shapiq/benchmark/configuration.py

@@ -34,7 +34,7 @@ information:
 """
 
 import os
-from typing import Any, Optional
+from typing import Any
 
 from shapiq.approximator import (
     FSII_APPROXIMATORS,
@@ -757,7 +757,7 @@ APPROXIMATION_BENCHMARK_PARAMS.update(
 
 
 def get_game_file_name_from_config(
-    configuration: dict[str, Any], iteration: Optional[int] = None
+    configuration: dict[str, Any], iteration: int | None = None
 ) -> str:
     """Get the file name for the game data with the given configuration and iteration.
 

{shapiq-1.2.1 → shapiq-1.2.2}/shapiq/benchmark/load.py

@@ -4,7 +4,7 @@ from the precomputed data (GitHub repository)."""
 import os
 import time
 from collections.abc import Generator
-from typing import Any, Optional, Union
+from typing import Any
 
 import requests
 
@@ -26,9 +26,9 @@ __all__ = [
 
 
 def load_games_from_configuration(
-    game_class: Union[Game.__class__, str],
+    game_class: Game.__class__ | str,
     config_id: int,
-    n_games: Optional[int] = None,
+    n_games: int | None = None,
     n_player_id: int = 0,
     check_pre_computed: bool = True,
     only_pre_computed: bool = True,

{shapiq-1.2.1 → shapiq-1.2.2}/shapiq/benchmark/metrics.py

@@ -2,7 +2,6 @@
 
 import copy
 import warnings
-from typing import Optional
 
 import numpy as np
 from scipy.stats import kendalltau
@@ -145,7 +144,7 @@ def compute_precision_at_k(
 def get_all_metrics(
     ground_truth: InteractionValues,
     estimated: InteractionValues,
-    order_indicator: Optional[str] = None,
+    order_indicator: str | None = None,
 ) -> dict:
     """Get all metrics for the interaction values.
 

{shapiq-1.2.1 → shapiq-1.2.2}/shapiq/benchmark/plot.py

@@ -1,7 +1,7 @@
 """This module contains the plotting utilities for the benchmark results."""
 
 from collections import defaultdict
-from typing import Callable, Optional, Union
+from collections.abc import Callable
 
 import numpy as np
 import pandas as pd
@@ -162,14 +162,14 @@ def agg_percentile(q: float) -> Callable[[np.ndarray], float]:
 
 
 def plot_approximation_quality(
-    data: Optional[pd.DataFrame] = None,
+    data: pd.DataFrame | None = None,
     *,
-    data_path: Optional[str] = None,
+    data_path: str | None = None,
     metric: str = "MSE",
-    orders: Optional[list[Union[int, str]]] = None,
-    approximators: Optional[list[str]] = None,
+    orders: list[int | str] | None = None,
+    approximators: list[str] | None = None,
     aggregation: str = "mean",
-    confidence_metric: Optional[str] = "sem",
+    confidence_metric: str | None = "sem",
     log_scale_y: bool = False,
     log_scale_min: float = LOG_SCALE_MIN,
     log_scale_max: float = LOG_SCALE_MAX,
@@ -421,8 +421,8 @@ def get_metric_data(results_df: pd.DataFrame, metric: str = "MSE") -> pd.DataFra
 
 def add_legend(
     axis: plt.Axes,
-    approximators: list[Union[str, Approximator]],
-    orders: Optional[list[Union[int, str]]] = None,
+    approximators: list[str | Approximator],
+    orders: list[int | str] | None = None,
     legend_subtitle: bool = True,
     loc: str = "best",
 ) -> None: