Add observable class

.pre-commit-config.yaml

@@ -4,17 +4,17 @@ fail_fast: true
 
 repos:
   - repo: https://github.com/psf/black
-    rev: 22.8.0
+    rev: 23.3.0
     hooks:
       - id: black
 
   - repo: https://github.com/timothycrosley/isort
-    rev: 5.10.1
+    rev: 5.12.0
     hooks:
       - id: isort
 
   - repo: https://github.com/pycqa/flake8
-    rev: 5.0.4
+    rev: 6.0.0
     hooks:
       - id: flake8
         additional_dependencies: [flake8-isort]

znnl/distance_metrics/__init__.py

@@ -26,6 +26,7 @@
 """
 from znnl.distance_metrics.angular_distance import AngularDistance
 from znnl.distance_metrics.cosine_distance import CosineDistance
+from znnl.distance_metrics.cross_entropy_distance import CrossEntropyDistance
 from znnl.distance_metrics.distance_metric import DistanceMetric
 from znnl.distance_metrics.hyper_sphere_distance import HyperSphere
 from znnl.distance_metrics.l_p_norm import LPNorm
@@ -40,4 +41,5 @@
     OrderNDifference.__name__,
     MahalanobisDistance.__name__,
     HyperSphere.__name__,
+    CrossEntropyDistance.__name__,
 ]

znnl/distance_metrics/angular_distance.py

@@ -50,6 +50,17 @@ def __init__(self, points: int = None):
         else:
             raise ValueError("Invalid points input.")
 
+    def __name__(self):
+        """
+        Name of the class.
+
+        Returns
+        -------
+        name : str
+                The name of the class.
+        """
+        return "angular_distance"
+
     def __call__(self, point_1: np.ndarray, point_2: np.ndarray, **kwargs):
         """
         Call the distance metric.

znnl/distance_metrics/cosine_distance.py

@@ -68,3 +68,14 @@ def __call__(self, point_1: np.ndarray, point_2: np.ndarray, **kwargs):
         )
 
         return 1 - abs(np.divide(numerator, denominator))
+
+    def __name__(self):
+        """
+        Name of the class.
+
+        Returns
+        -------
+        name : str
+                The name of the class.
+        """
+        return "cosine_distance"

znnl/distance_metrics/cross_entropy_distance.py

@@ -0,0 +1,61 @@
+"""
+ZnNL: A Zincwarecode package.
+
+License
+-------
+This program and the accompanying materials are made available under the terms
+of the Eclipse Public License v2.0 which accompanies this distribution, and is
+available at https://www.eclipse.org/legal/epl-v20.html
+
+SPDX-License-Identifier: EPL-2.0
+
+Copyright Contributors to the Zincwarecode Project.
+
+Contact Information
+-------------------
+email: zincwarecode@gmail.com
+github: https://github.com/zincware
+web: https://zincwarecode.com/
+
+Citation
+--------
+If you use this module please cite us with:
+
+Summary
+-------
+"""
+import optax
+
+from znnl.distance_metrics.distance_metric import DistanceMetric
+
+
+class CrossEntropyDistance(DistanceMetric):
+    """
+    Class for the cross entropy distance
+    """
+
+    def __name__(self):
+        """
+        Name of the class.
+
+        Returns
+        -------
+        name : str
+                The name of the class.
+        """
+        return "cross_entropy_distance"
+
+    def __call__(self, prediction, target):
+        """
+
+        Parameters
+        ----------
+        prediction (batch_size, n_classes)
+        target
+
+        Returns
+        -------
+        Softmax cross entropy of the batch.
+
+        """
+        return optax.softmax_cross_entropy(logits=prediction, labels=target)

znnl/distance_metrics/distance_metric.py

@@ -26,12 +26,37 @@
 """
 import jax.numpy as np
 
+from znnl.observables.observable import Observable
 
-class DistanceMetric:
+
+class DistanceMetric(Observable):
     """
     Parent class for a ZnRND distance metric.
     """
 
+    def __name__(self) -> str:
+        """
+        Name of the class.
+
+        Returns
+        -------
+        name : str
+                The name of the class.
+        """
+        return "distance_metric"
+
+    def __signature__(self) -> tuple:
+        """
+        Signature of the class.
+
+        Returns
+        -------
+        signature : tuple
+                The signature of the class.
+                For the distance metric, it is (1,).
+        """
+        return (1,)
+
     def __call__(self, point_1: np.ndarray, point_2: np.ndarray, **kwargs):
         """
         Call the distance metric.

znnl/distance_metrics/hyper_sphere_distance.py

@@ -73,3 +73,14 @@ def __call__(self, point_1: np.ndarray, point_2: np.ndarray, **kwargs):
         return LPNorm(order=self.order)(point_1, point_2) * CosineDistance()(
             point_1, point_2
         )
+
+    def __name__(self):
+        """
+        Name of the class.
+
+        Returns
+        -------
+        name : str
+                The name of the class.
+        """
+        return f"hyper_sphere_distance_{self.order}"

znnl/distance_metrics/l_p_norm.py

@@ -69,3 +69,14 @@ def __call__(self, point_1: np.ndarray, point_2: np.ndarray, **kwargs):
                 Array of distances for each point.
         """
         return np.linalg.norm(point_1 - point_2, axis=1, ord=self.order)
+
+    def __name__(self):
+        """
+        Name of the class.
+
+        Returns
+        -------
+        name : str
+                The name of the class.
+        """
+        return f"lp_norm_{self.order}"

znnl/distance_metrics/mahalanobis_distance.py

@@ -66,3 +66,14 @@ def __call__(self, point_1: np.array, point_2: np.array, **kwargs) -> np.array:
             distances.append(distance)
 
         return distances
+
+    def __name__(self):
+        """
+        Name of the class.
+
+        Returns
+        -------
+        name : str
+                The name of the class.
+        """
+        return "mahalanobis_distance"

znnl/distance_metrics/order_n_difference.py

@@ -79,3 +79,14 @@ def __call__(self, point_1: np.ndarray, point_2: np.ndarray, **kwargs):
             return np.sum(np.power(diff, self.order), axis=1)
         else:
             raise ValueError(f"Invalid reduction operation: {self.reduce_operation}")
+
+    def __name__(self):
+        """
+        Name of the class.
+
+        Returns
+        -------
+        name : str
+                The name of the class.
+        """
+        return "order_{self.order}_difference_{self.reduce_operation}"

znnl/loss_functions/__init__.py

@@ -28,16 +28,16 @@
 from znnl.loss_functions.cosine_distance import CosineDistanceLoss
 from znnl.loss_functions.cross_entropy_loss import CrossEntropyLoss
 from znnl.loss_functions.l_p_norm import LPNormLoss
+from znnl.loss_functions.loss import Loss
 from znnl.loss_functions.mahalanobis import MahalanobisLoss
 from znnl.loss_functions.mean_power_error import MeanPowerLoss
-from znnl.loss_functions.simple_loss import SimpleLoss
 
 __all__ = [
     AngleDistanceLoss.__name__,
     CosineDistanceLoss.__name__,
     LPNormLoss.__name__,
     MahalanobisLoss.__name__,
     MeanPowerLoss.__name__,
-    SimpleLoss.__name__,
+    Loss.__name__,
     CrossEntropyLoss.__name__,
 ]

znnl/loss_functions/absolute_angle_difference.py

@@ -25,10 +25,10 @@
 -------
 """
 from znnl.distance_metrics.angular_distance import AngularDistance
-from znnl.loss_functions.simple_loss import SimpleLoss
+from znnl.loss_functions.loss import Loss
 
 
-class AngleDistanceLoss(SimpleLoss):
+class AngleDistanceLoss(Loss):
     """
     Class for the mean power loss
     """
@@ -39,3 +39,14 @@ def __init__(self):
         """
         super(AngleDistanceLoss, self).__init__()
         self.metric = AngularDistance()
+
+    def __name__(self):
+        """
+        Name of the class.
+
+        Returns
+        -------
+        name : str
+                The name of the class.
+        """
+        return "angle_distance_loss"

znnl/loss_functions/cosine_distance.py

@@ -25,10 +25,10 @@
 -------
 """
 from znnl.distance_metrics.cosine_distance import CosineDistance
-from znnl.loss_functions.simple_loss import SimpleLoss
+from znnl.loss_functions.loss import Loss
 
 
-class CosineDistanceLoss(SimpleLoss):
+class CosineDistanceLoss(Loss):
     """
     Class for the mean power loss
     """
@@ -39,3 +39,14 @@ def __init__(self):
         """
         super(CosineDistanceLoss, self).__init__()
         self.metric = CosineDistance()
+
+    def __name__(self):
+        """
+        Name of the class.
+
+        Returns
+        -------
+        name : str
+                The name of the class.
+        """
+        return "cosine_distance_loss"

znnl/loss_functions/cross_entropy_loss.py

@@ -24,33 +24,11 @@
 Summary
 -------
 """
-import optax
+from znnl.distance_metrics.cross_entropy_distance import CrossEntropyDistance
+from znnl.loss_functions.loss import Loss
 
-from znnl.loss_functions.simple_loss import SimpleLoss
 
-
-class CrossEntropyDistance:
-    """
-    Class for the cross entropy distance
-    """
-
-    def __call__(self, prediction, target):
-        """
-
-        Parameters
-        ----------
-        prediction (batch_size, n_classes)
-        target
-
-        Returns
-        -------
-        Softmax cross entropy of the batch.
-
-        """
-        return optax.softmax_cross_entropy(logits=prediction, labels=target)
-
-
-class CrossEntropyLoss(SimpleLoss):
+class CrossEntropyLoss(Loss):
     """
     Class for the cross entropy loss
     """
@@ -61,3 +39,14 @@ def __init__(self):
         """
         super(CrossEntropyLoss, self).__init__()
         self.metric = CrossEntropyDistance()
+
+    def __name__(self):
+        """
+        Name of the class.
+
+        Returns
+        -------
+        name : str
+                The name of the class.
+        """
+        return "cross_entropy_loss"

znnl/loss_functions/l_p_norm.py

@@ -25,7 +25,7 @@
 -------
 """
 from znnl.distance_metrics.l_p_norm import LPNorm
-from znnl.loss_functions.simple_loss import SimpleLoss
+from znnl.loss_functions.loss import SimpleLoss
 
 
 class LPNormLoss(SimpleLoss):
@@ -44,3 +44,14 @@ def __init__(self, order: float):
         """
         super(LPNormLoss, self).__init__()
         self.metric = LPNorm(order=order)
+
+    def __name__(self):
+        """
+        Name of the class.
+
+        Returns
+        -------
+        name : str
+                The name of the class.
+        """
+        return f"lp_norm_loss_{self.order}"