Merge pull request #2 from si-cim/new-components

Create Component and initializer classes.

examples/new_components.py

@@ -0,0 +1,67 @@
+#
+# DATASET
+#
+import torch
+
+from sklearn.datasets import load_iris
+from sklearn.preprocessing import StandardScaler
+
+scaler = StandardScaler()
+x_train, y_train = load_iris(return_X_y=True)
+x_train = x_train[:, [0, 2]]
+scaler.fit(x_train)
+x_train = scaler.transform(x_train)
+
+x_train = torch.Tensor(x_train)
+y_train = torch.Tensor(y_train)
+num_classes = len(torch.unique(y_train))
+
+#
+# CREATE NEW COMPONENTS
+#
+from prototorch.components import *
+from prototorch.components.initializers import *
+
+unsupervised = Components(6, SelectionInitializer(x_train))
+print(unsupervised())
+
+prototypes = LabeledComponents(
+    (3, 2), StratifiedSelectionInitializer(x_train, y_train))
+print(prototypes())
+
+components = ReasoningComponents(
+    (3, 6), StratifiedSelectionInitializer(x_train, y_train))
+print(components())
+
+#
+# TEST SERIALIZATION
+#
+import io
+
+save = io.BytesIO()
+torch.save(unsupervised, save)
+save.seek(0)
+serialized_unsupervised = torch.load(save)
+
+assert torch.all(unsupervised.components == serialized_unsupervised.components
+                 ), "Serialization of Components failed."
+
+save = io.BytesIO()
+torch.save(prototypes, save)
+save.seek(0)
+serialized_prototypes = torch.load(save)
+
+assert torch.all(prototypes.components == serialized_prototypes.components
+                 ), "Serialization of Components failed."
+assert torch.all(prototypes.labels == serialized_prototypes.labels
+                 ), "Serialization of Components failed."
+
+save = io.BytesIO()
+torch.save(components, save)
+save.seek(0)
+serialized_components = torch.load(save)
+
+assert torch.all(components.components == serialized_components.components
+                 ), "Serialization of Components failed."
+assert torch.all(components.reasonings == serialized_components.reasonings
+                 ), "Serialization of Components failed."

prototorch/components/__init__.py

@@ -0,0 +1,7 @@
+from prototorch.components.components import Components, LabeledComponents, ReasoningComponents
+
+__all__ = [
+    "Components",
+    "LabeledComponents",
+    "ReasoningComponents",
+]

prototorch/components/components.py

@@ -0,0 +1,130 @@
+"""ProtoTorch components modules."""
+
+from typing import Tuple
+import warnings
+from prototorch.components.initializers import EqualLabelInitializer, ZeroReasoningsInitializer
+import torch
+from torch.nn.parameter import Parameter
+
+from prototorch.functions.initializers import get_initializer
+
+
+class Components(torch.nn.Module):
+    """
+    Components is a set of learnable Tensors.
+    """
+    def __init__(self,
+                 number_of_components=None,
+                 initializer=None,
+                 *,
+                 initialized_components=None,
+                 dtype=torch.float32):
+        super().__init__()
+
+        # Ignore all initialization settings if initialized_components is given.
+        if initialized_components is not None:
+            self._components = Parameter(initialized_components)
+            if number_of_components is not None or initializer is not None:
+                warnings.warn(
+                    "Arguments ignored while initializing Components")
+        else:
+            self._initialize_components(number_of_components, initializer)
+
+    def _initialize_components(self, number_of_components, initializer):
+        self._components = Parameter(
+            initializer.generate(number_of_components))
+
+    @property
+    def components(self):
+        """
+        Tensor containing the component tensors.
+        """
+        return self._components.detach().cpu()
+
+    def forward(self):
+        return self._components
+
+    def extra_repr(self):
+        return f"components.shape: {tuple(self._components.shape)}"
+
+
+class LabeledComponents(Components):
+    """
+    LabeledComponents generate a set of components and a set of labels.
+    Every Component has a label assigned.
+    """
+    def __init__(self,
+                 labels=None,
+                 initializer=None,
+                 *,
+                 initialized_components=None):
+        if initialized_components is not None:
+            super().__init__(initialized_components=initialized_components[0])
+            self._labels = initialized_components[1]
+        else:
+            self._initialize_labels(labels, initializer)
+            super().__init__(number_of_components=len(self._labels),
+                             initializer=initializer)
+
+    def _initialize_labels(self, labels, initializer):
+        if type(labels) == tuple:
+            num_classes, prototypes_per_class = labels
+            labels = EqualLabelInitializer(num_classes, prototypes_per_class)
+
+        self._labels = labels.generate()
+
+    @property
+    def labels(self):
+        """
+        Tensor containing the component tensors.
+        """
+        return self._labels.detach().cpu()
+
+    def forward(self):
+        return super().forward(), self._labels
+
+
+class ReasoningComponents(Components):
+    """
+    ReasoningComponents generate a set of components and a set of reasoning matrices.
+    Every Component has a reasoning matrix assigned.
+
+    A reasoning matrix is a Nx2 matrix, where N is the number of Classes.
+    The first element is called positive reasoning :math:`p`, the second negative reasoning :math:`n`.
+    A components can reason in favour (positive) of a class, against (negative) a class or not at all (neutral).
+
+    It holds that :math:`0 \leq n \leq 1`, :math:`0 \leq p \leq 1` and :math:`0 \leq n+p \leq 1`.
+    Therefore :math:`n` and  :math:`p` are two elements of a three element probability distribution.
+    """
+    def __init__(self,
+                 reasonings=None,
+                 initializer=None,
+                 *,
+                 initialized_components=None):
+        if initialized_components is not None:
+            super().__init__(initialized_components=initialized_components[0])
+            self._reasonings = initialized_components[1]
+        else:
+            self._initialize_reasonings(reasonings)
+            super().__init__(number_of_components=len(self._reasonings),
+                             initializer=initializer)
+
+    def _initialize_reasonings(self, reasonings):
+        if type(reasonings) == tuple:
+            num_classes, number_of_components = reasonings
+            reasonings = ZeroReasoningsInitializer(num_classes,
+                                                   number_of_components)
+
+        self._reasonings = reasonings.generate()
+
+    @property
+    def reasonings(self):
+        """
+        Returns Reasoning Matrix.
+
+        Dimension NxCx2
+        """
+        return self._reasonings.detach().cpu()
+
+    def forward(self):
+        return super().forward(), self._reasonings

prototorch/components/initializers.py

@@ -0,0 +1,132 @@
+import torch
+from collections.abc import Iterable
+
+
+# Components
+class ComponentsInitializer:
+    def generate(self, number_of_components):
+        pass
+
+
+class DimensionAwareInitializer(ComponentsInitializer):
+    def __init__(self, c_dims):
+        super().__init__()
+        if isinstance(c_dims, Iterable):
+            self.components_dims = tuple(c_dims)
+        else:
+            self.components_dims = (c_dims, )
+
+
+class OnesInitializer(DimensionAwareInitializer):
+    def generate(self, length):
+        gen_dims = (length, ) + self.components_dims
+        return torch.ones(gen_dims)
+
+
+class ZerosInitializer(DimensionAwareInitializer):
+    def generate(self, length):
+        gen_dims = (length, ) + self.components_dims
+        return torch.zeros(gen_dims)
+
+
+class UniformInitializer(DimensionAwareInitializer):
+    def __init__(self, c_dims, min=0.0, max=1.0):
+        super().__init__(c_dims)
+
+        self.min = min
+        self.max = max
+
+    def generate(self, length):
+        gen_dims = (length, ) + self.components_dims
+        return torch.FloatTensor(gen_dims).uniform_(self.min, self.max)
+
+
+class PositionAwareInitializer(ComponentsInitializer):
+    def __init__(self, positions):
+        super().__init__()
+        self.data = positions
+
+
+class SelectionInitializer(PositionAwareInitializer):
+    def generate(self, length):
+        indices = torch.LongTensor(length).random_(0, len(self.data))
+        return self.data[indices]
+
+
+class MeanInitializer(PositionAwareInitializer):
+    def generate(self, length):
+        mean = torch.mean(self.data, dim=0)
+        repeat_dim = [length] + [1] * len(mean.shape)
+        return mean.repeat(repeat_dim)
+
+
+class ClassAwareInitializer(ComponentsInitializer):
+    def __init__(self, positions, classes):
+        super().__init__()
+        self.data = positions
+        self.classes = classes
+
+        self.names = torch.unique(self.classes)
+        self.num_classes = len(self.names)
+
+
+class StratifiedMeanInitializer(ClassAwareInitializer):
+    def __init__(self, positions, classes):
+        super().__init__(positions, classes)
+
+        self.initializers = []
+        for name in self.names:
+            class_data = self.data[self.classes == name]
+            class_initializer = MeanInitializer(class_data)
+            self.initializers.append(class_initializer)
+
+    def generate(self, length):
+        per_class = length // self.num_classes
+        return torch.vstack(
+            [init.generate(per_class) for init in self.initializers])
+
+
+class StratifiedSelectionInitializer(ClassAwareInitializer):
+    def __init__(self, positions, classes):
+        super().__init__(positions, classes)
+
+        self.initializers = []
+        for name in self.names:
+            class_data = self.data[self.classes == name]
+            class_initializer = SelectionInitializer(class_data)
+            self.initializers.append(class_initializer)
+
+    def generate(self, length):
+        per_class = length // self.num_classes
+        return torch.vstack(
+            [init.generate(per_class) for init in self.initializers])
+
+
+# Labels
+class LabelsInitializer:
+    def generate(self):
+        pass
+
+
+class EqualLabelInitializer(LabelsInitializer):
+    def __init__(self, classes, per_class):
+        self.classes = classes
+        self.per_class = per_class
+
+    def generate(self):
+        return torch.arange(self.classes).repeat(self.per_class, 1).T.flatten()
+
+
+# Reasonings
+class ReasoningsInitializer:
+    def generate(self, length):
+        pass
+
+
+class ZeroReasoningsInitializer(ReasoningsInitializer):
+    def __init__(self, classes, length):
+        self.classes = classes
+        self.length = length
+
+    def generate(self):
+        return torch.zeros((self.length, self.classes, 2))

prototorch/modules/prototypes.py

@@ -39,6 +39,9 @@ class Prototypes1D(_Prototypes):
         one_hot_labels=False,
         **kwargs,
     ):
+        warnings.warn(
+            PendingDeprecationWarning(
+                "Prototypes1D will be replaced in future versions."))
 
         # Convert tensors to python lists before processing
         if prototype_distribution is not None: