Add scaffolding

Based on Soft LVQ paper by Seo and Obermayer

.bumpversion.cfg

@@ -1,5 +1,5 @@
 [bumpversion]
-current_version = 0.4.1
+current_version = 0.5.0
 commit = True
 tag = True
 parse = (?P<major>\d+)\.(?P<minor>\d+)\.(?P<patch>\d+)

.github/ISSUE_TEMPLATE/bug_report.md

@@ -23,9 +23,9 @@ A clear and concise description of what you expected to happen.
 If applicable, add screenshots to help explain your problem.
 
 **Desktop (please complete the following information):**
- - OS: [e.g. Ubuntu 20.10]
+- OS: [e.g. Ubuntu 20.10]
- - Prototorch Version: [e.g. v0.4.0]
+- Prototorch Version: [e.g. v0.4.0]
- - Python Version: [e.g. 3.9.5]
+- Python Version: [e.g. 3.9.5]
 
 **Additional context**
-Add any other context about the problem here.
+Add any other context about the problem here.

.gitignore

@@ -154,4 +154,5 @@ scratch*
 # End of https://www.gitignore.io/api/visualstudiocode
 .vscode/
 
-reports
+reports
+artifacts

.travis.yml

@@ -4,7 +4,9 @@ language: python
 python: 3.8
 cache:
   directories:
+  - "$HOME/.cache/pip"
   - "./tests/artifacts"
+  - "$HOME/datasets"
 install:
 - pip install .[all] --progress-bar off
 

RELEASE.md

@@ -1,13 +1,16 @@
 # ProtoTorch Releases
 
+## Release 0.5.0
+
+- Breaking: Removed deprecated `prototorch.modules.Prototypes1D`.
+  - Use `prototorch.components.LabeledComponents` instead.
+
 ## Release 0.2.0
 
-### Includes
 - Fixes in example scripts.
 
 ## Release 0.1.1-dev0
 
-### Includes
 - Minor bugfixes.
 - 100% line coverage.
 

docs/source/api.rst

@@ -1,13 +1,24 @@
-.. ProtoFlow API Reference
+.. ProtoTorch API Reference
 
-ProtoFlow API Reference
+ProtoTorch API Reference
 ======================================
 
 Datasets
 --------------------------------------
+
+Common Datasets
+~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
 .. automodule:: prototorch.datasets
    :members:
-   :undoc-members:
+
+
+Abstract Datasets
+~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
+
+Abstract Datasets are used to build your own datasets.
+
+.. autoclass:: prototorch.datasets.abstract.NumpyDataset
+   :members:
 
 Functions
 --------------------------------------

docs/source/conf.py

@@ -23,7 +23,7 @@ author = "Jensun Ravichandran"
 
 # The full version, including alpha/beta/rc tags
 #
-release = "0.4.1"
+release = "0.5.0"
 
 # -- General configuration ---------------------------------------------------
 
@@ -46,6 +46,7 @@ extensions = [
     "sphinx.ext.viewcode",
     "sphinx_rtd_theme",
     "sphinxcontrib.katex",
+    'sphinx_autodoc_typehints',
 ]
 
 # katex_prerender = True
@@ -179,6 +180,9 @@ texinfo_documents = [
 intersphinx_mapping = {
     "python": ("https://docs.python.org/", None),
     "numpy": ("https://docs.scipy.org/doc/numpy/", None),
+    "torch": ('https://pytorch.org/docs/stable/', None),
+    "pytorch_lightning":
+    ("https://pytorch-lightning.readthedocs.io/en/stable/", None),
 }
 
 # -- Options for Epub output ----------------------------------------------

examples/glvq_iris.py

@@ -3,14 +3,13 @@
 import numpy as np
 import torch
 from matplotlib import pyplot as plt
-from sklearn.datasets import load_iris
+from prototorch.components import LabeledComponents, StratifiedMeanInitializer
-from sklearn.preprocessing import StandardScaler
-from torchinfo import summary
-
 from prototorch.functions.competitions import wtac
 from prototorch.functions.distances import euclidean_distance
 from prototorch.modules.losses import GLVQLoss
-from prototorch.modules.prototypes import Prototypes1D
+from sklearn.datasets import load_iris
+from sklearn.preprocessing import StandardScaler
+from torchinfo import summary
 
 # Prepare and preprocess the data
 scaler = StandardScaler()
@@ -25,19 +24,17 @@ class Model(torch.nn.Module):
     def __init__(self):
         """GLVQ model for training on 2D Iris data."""
         super().__init__()
-        self.proto_layer = Prototypes1D(
+        prototype_initializer = StratifiedMeanInitializer([x_train, y_train])
-            input_dim=2,
+        prototype_distribution = {"num_classes": 3, "prototypes_per_class": 3}
-            prototypes_per_class=3,
+        self.proto_layer = LabeledComponents(
-            nclasses=3,
+            prototype_distribution,
-            prototype_initializer="stratified_random",
+            prototype_initializer,
-            data=[x_train, y_train],
         )
 
     def forward(self, x):
-        protos = self.proto_layer.prototypes
+        prototypes, prototype_labels = self.proto_layer()
-        plabels = self.proto_layer.prototype_labels
+        distances = euclidean_distance(x, prototypes)
-        dis = euclidean_distance(x, protos)
+        return distances, prototype_labels
-        return dis, plabels
 
 
 # Build the GLVQ model
@@ -54,43 +51,46 @@ x_in = torch.Tensor(x_train)
 y_in = torch.Tensor(y_train)
 
 # Training loop
-title = "Prototype Visualization"
+TITLE = "Prototype Visualization"
-fig = plt.figure(title)
+fig = plt.figure(TITLE)
 for epoch in range(70):
     # Compute loss
-    dis, plabels = model(x_in)
+    distances, prototype_labels = model(x_in)
-    loss = criterion([dis, plabels], y_in)
+    loss = criterion([distances, prototype_labels], y_in)
+
+    # Compute Accuracy
     with torch.no_grad():
-        pred = wtac(dis, plabels)
+        predictions = wtac(distances, prototype_labels)
-        correct = pred.eq(y_in.view_as(pred)).sum().item()
+        correct = predictions.eq(y_in.view_as(predictions)).sum().item()
     acc = 100.0 * correct / len(x_train)
+
     print(
         f"Epoch: {epoch + 1:03d} Loss: {loss.item():05.02f} Acc: {acc:05.02f}%"
     )
 
-    # Take a gradient descent step
+    # Optimizer step
     optimizer.zero_grad()
     loss.backward()
     optimizer.step()
 
     # Get the prototypes form the model
-    protos = model.proto_layer.prototypes.data.numpy()
+    prototypes = model.proto_layer.components.numpy()
-    if np.isnan(np.sum(protos)):
+    if np.isnan(np.sum(prototypes)):
         print("Stopping training because of `nan` in prototypes.")
         break
 
     # Visualize the data and the prototypes
     ax = fig.gca()
     ax.cla()
-    ax.set_title(title)
+    ax.set_title(TITLE)
     ax.set_xlabel("Data dimension 1")
     ax.set_ylabel("Data dimension 2")
     cmap = "viridis"
     ax.scatter(x_train[:, 0], x_train[:, 1], c=y_train, edgecolor="k")
     ax.scatter(
-        protos[:, 0],
+        prototypes[:, 0],
-        protos[:, 1],
+        prototypes[:, 1],
-        c=plabels,
+        c=prototype_labels,
         cmap=cmap,
         edgecolor="k",
         marker="D",
@@ -98,7 +98,7 @@ for epoch in range(70):
     )
 
     # Paint decision regions
-    x = np.vstack((x_train, protos))
+    x = np.vstack((x_train, prototypes))
     x_min, x_max = x[:, 0].min() - 1, x[:, 0].max() + 1
     y_min, y_max = x[:, 1].min() - 1, x[:, 1].max() + 1
     xx, yy = np.meshgrid(np.arange(x_min, x_max, 1 / 50),
@@ -108,7 +108,7 @@ for epoch in range(70):
     torch_input = torch.Tensor(mesh_input)
     d = model(torch_input)[0]
     w_indices = torch.argmin(d, dim=1)
-    y_pred = torch.index_select(plabels, 0, w_indices)
+    y_pred = torch.index_select(prototype_labels, 0, w_indices)
     y_pred = y_pred.reshape(xx.shape)
 
     # Plot voronoi regions

examples/gmlvq_tecator.py

@@ -2,13 +2,12 @@
 
 import matplotlib.pyplot as plt
 import torch
-from torch.utils.data import DataLoader
+from prototorch.components import LabeledComponents, StratifiedMeanInitializer
-
 from prototorch.datasets.tecator import Tecator
 from prototorch.functions.distances import sed
-from prototorch.modules import Prototypes1D
 from prototorch.modules.losses import GLVQLoss
 from prototorch.utils.colors import get_legend_handles
+from torch.utils.data import DataLoader
 
 # Prepare the dataset and dataloader
 train_data = Tecator(root="./artifacts", train=True)
@@ -19,22 +18,22 @@ class Model(torch.nn.Module):
     def __init__(self, **kwargs):
         """GMLVQ model as a siamese network."""
         super().__init__()
-        x, y = train_data.data, train_data.targets
+        prototype_initializer = StratifiedMeanInitializer(train_loader)
-        self.p1 = Prototypes1D(
+        prototype_distribution = {"num_classes": 2, "prototypes_per_class": 2}
-            input_dim=100,
+
-            prototypes_per_class=2,
+        self.proto_layer = LabeledComponents(
-            nclasses=2,
+            prototype_distribution,
-            prototype_initializer="stratified_random",
+            prototype_initializer,
-            data=[x, y],
         )
+
         self.omega = torch.nn.Linear(in_features=100,
                                      out_features=100,
                                      bias=False)
         torch.nn.init.eye_(self.omega.weight)
 
     def forward(self, x):
-        protos = self.p1.prototypes
+        protos = self.proto_layer.components
-        plabels = self.p1.prototype_labels
+        plabels = self.proto_layer.component_labels
 
         # Process `x` and `protos` through `omega`
         x_map = self.omega(x)
@@ -86,8 +85,8 @@ im = ax.imshow(omega.dot(omega.T), cmap="viridis")
 plt.show()
 
 # Get the prototypes form the model
-protos = model.p1.prototypes.data.numpy()
+protos = model.proto_layer.components.numpy()
-plabels = model.p1.prototype_labels
+plabels = model.proto_layer.component_labels.numpy()
 
 # Visualize the prototypes
 title = "Tecator Prototypes"

examples/gtlvq_mnist.py

@@ -12,20 +12,19 @@ import numpy as np
 import torch
 import torch.nn as nn
 import torchvision
-from torchvision import transforms
-
 from prototorch.functions.helper import calculate_prototype_accuracy
 from prototorch.modules.losses import GLVQLoss
 from prototorch.modules.models import GTLVQ
+from torchvision import transforms
 
 # Parameters and options
-n_epochs = 50
+num_epochs = 50
 batch_size_train = 64
 batch_size_test = 1000
 learning_rate = 0.1
 momentum = 0.5
 log_interval = 10
-cuda = "cuda:1"
+cuda = "cuda:0"
 random_seed = 1
 device = torch.device(cuda if torch.cuda.is_available() else "cpu")
 
@@ -141,14 +140,14 @@ optimizer = torch.optim.Adam(
 criterion = GLVQLoss(squashing="sigmoid_beta", beta=10)
 
 # Training loop
-for epoch in range(n_epochs):
+for epoch in range(num_epochs):
     for batch_idx, (x_train, y_train) in enumerate(train_loader):
         model.train()
         x_train, y_train = x_train.to(device), y_train.to(device)
         optimizer.zero_grad()
 
         distances = model(x_train)
-        plabels = model.gtlvq.cls.prototype_labels.to(device)
+        plabels = model.gtlvq.cls.component_labels.to(device)
 
         # Compute loss.
         loss = criterion([distances, plabels], y_train)
@@ -161,7 +160,7 @@ for epoch in range(n_epochs):
         if batch_idx % log_interval == 0:
             acc = calculate_prototype_accuracy(distances, y_train, plabels)
             print(
-                f"Epoch: {epoch + 1:02d}/{n_epochs:02d} Epoch Progress: {100. * batch_idx / len(train_loader):02.02f} % Loss: {loss.item():02.02f} \
+                f"Epoch: {epoch + 1:02d}/{num_epochs:02d} Epoch Progress: {100. * batch_idx / len(train_loader):02.02f} % Loss: {loss.item():02.02f} \
               Train Acc: {acc.item():02.02f}")
 
     # Test

examples/lgmlvq_iris.py

@@ -3,14 +3,12 @@
 import numpy as np
 import torch
 from matplotlib import pyplot as plt
-from sklearn.datasets import load_iris
+from prototorch.components import LabeledComponents, StratifiedMeanInitializer
-from sklearn.metrics import accuracy_score
-
 from prototorch.functions.competitions import stratified_min
 from prototorch.functions.distances import lomega_distance
-from prototorch.functions.init import eye_
 from prototorch.modules.losses import GLVQLoss
-from prototorch.modules.prototypes import Prototypes1D
+from sklearn.datasets import load_iris
+from sklearn.metrics import accuracy_score
 
 # Prepare training data
 x_train, y_train = load_iris(True)
@@ -22,19 +20,19 @@ class Model(torch.nn.Module):
     def __init__(self):
         """Local-GMLVQ model."""
         super().__init__()
-        self.p1 = Prototypes1D(
+
-            input_dim=2,
+        prototype_initializer = StratifiedMeanInitializer([x_train, y_train])
-            prototype_distribution=[1, 2, 2],
+        prototype_distribution = [1, 2, 2]
-            prototype_initializer="stratified_random",
+        self.proto_layer = LabeledComponents(
-            data=[x_train, y_train],
+            prototype_distribution,
+            prototype_initializer,
         )
-        omegas = torch.zeros(5, 2, 2)
+
+        omegas = torch.eye(2, 2).repeat(5, 1, 1)
         self.omegas = torch.nn.Parameter(omegas)
-        eye_(self.omegas)
 
     def forward(self, x):
-        protos = self.p1.prototypes
+        protos, plabels = self.proto_layer()
-        plabels = self.p1.prototype_labels
         omegas = self.omegas
         dis = lomega_distance(x, protos, omegas)
         return dis, plabels
@@ -69,7 +67,7 @@ for epoch in range(100):
     optimizer.step()
 
     # Get the prototypes form the model
-    protos = model.p1.prototypes.data.numpy()
+    protos = model.proto_layer.components.numpy()
 
     # Visualize the data and the prototypes
     ax = fig.gca()

prototorch/__init__.py

@@ -1,21 +1,24 @@
 """ProtoTorch package."""
 
+import pkgutil
+
+import pkg_resources
+
+from . import components, datasets, functions, modules, utils
+from .datasets import *
+
 # Core Setup
-__version__ = "0.4.1"
+__version__ = "0.5.0"
 
 __all_core__ = [
     "datasets",
     "functions",
     "modules",
+    "components",
+    "utils",
 ]
 
-from .datasets import *
-
 # Plugin Loader
-import pkgutil
-
-import pkg_resources
-
 __path__ = pkgutil.extend_path(__path__, __name__)
 
 

prototorch/components/components.py

@@ -1,54 +1,107 @@
 """ProtoTorch components modules."""
 
 import warnings
-from typing import Tuple
 
 import torch
-from prototorch.components.initializers import (ComponentsInitializer,
+from prototorch.components.initializers import (ClassAwareInitializer,
-                                                EqualLabelInitializer,
+                                                ComponentsInitializer,
+                                                CustomLabelsInitializer,
+                                                EqualLabelsInitializer,
+                                                UnequalLabelsInitializer,
                                                 ZeroReasoningsInitializer)
-from prototorch.functions.initializers import get_initializer
 from torch.nn.parameter import Parameter
 
+from .initializers import parse_data_arg
+
+
+def get_labels_object(distribution):
+    if isinstance(distribution, dict):
+        if "num_classes" in distribution.keys():
+            labels = EqualLabelsInitializer(
+                distribution["num_classes"],
+                distribution["prototypes_per_class"])
+        else:
+            labels = CustomLabelsInitializer(distribution)
+    elif isinstance(distribution, tuple):
+        num_classes, prototypes_per_class = distribution
+        labels = EqualLabelsInitializer(num_classes, prototypes_per_class)
+    elif isinstance(distribution, list):
+        labels = UnequalLabelsInitializer(distribution)
+    else:
+        msg = f"`distribution` not understood." \
+            f"You have provided: {distribution=}."
+        raise ValueError(msg)
+    return labels
+
+
+def _precheck_initializer(initializer):
+    if not isinstance(initializer, ComponentsInitializer):
+        emsg = f"`initializer` has to be some subtype of " \
+            f"{ComponentsInitializer}. " \
+            f"You have provided: {initializer=} instead."
+        raise TypeError(emsg)
+
 
 class Components(torch.nn.Module):
     """Components is a set of learnable Tensors."""
     def __init__(self,
-                 number_of_components=None,
+                 num_components=None,
                  initializer=None,
                  *,
-                 initialized_components=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)
+            self._register_components(initialized_components)
-            if number_of_components is not None or initializer is not None:
+            if num_components is not None or initializer is not None:
                 wmsg = "Arguments ignored while initializing Components"
                 warnings.warn(wmsg)
         else:
-            self._initialize_components(number_of_components, initializer)
+            self._initialize_components(num_components, initializer)
 
-    def _initialize_components(self, number_of_components, initializer):
+    @property
-        if not isinstance(initializer, ComponentsInitializer):
+    def num_components(self):
-            emsg = f"`initializer` has to be some subtype of " \
+        return len(self._components)
-                f"{ComponentsInitializer}. " \
+
-                f"You have provided: {initializer=} instead."
+    def _register_components(self, components):
-            raise TypeError(emsg)
+        self.register_parameter("_components", Parameter(components))
-        self._components = Parameter(
+
-            initializer.generate(number_of_components))
+    def _initialize_components(self, num_components, initializer):
+        _precheck_initializer(initializer)
+        _components = initializer.generate(num_components)
+        self._register_components(_components)
+
+    def add_components(self,
+                       num=1,
+                       initializer=None,
+                       *,
+                       initialized_components=None):
+        if initialized_components is not None:
+            _components = torch.cat([self._components, initialized_components])
+        else:
+            _precheck_initializer(initializer)
+            _new = initializer.generate(num)
+            _components = torch.cat([self._components, _new])
+        self._register_components(_components)
+
+    def remove_components(self, indices=None):
+        mask = torch.ones(self.num_components, dtype=torch.bool)
+        mask[indices] = False
+        _components = self._components[mask]
+        self._register_components(_components)
+        return mask
 
     @property
     def components(self):
         """Tensor containing the component tensors."""
-        return self._components.detach().cpu()
+        return self._components.detach()
 
     def forward(self):
         return self._components
 
     def extra_repr(self):
-        return f"components.shape: {tuple(self._components.shape)}"
+        return f"(components): (shape: {tuple(self._components.shape)})"
 
 
 class LabeledComponents(Components):
@@ -57,29 +110,70 @@ class LabeledComponents(Components):
     Every Component has a label assigned.
     """
     def __init__(self,
-                 labels=None,
+                 distribution=None,
                  initializer=None,
                  *,
                  initialized_components=None):
         if initialized_components is not None:
-            super().__init__(initialized_components=initialized_components[0])
+            components, component_labels = parse_data_arg(
-            self._labels = initialized_components[1]
+                initialized_components)
+            super().__init__(initialized_components=components)
+            self._labels = component_labels
         else:
-            self._initialize_labels(labels)
+            labels = get_labels_object(distribution)
-            super().__init__(number_of_components=len(self._labels),
+            self.initial_distribution = labels.distribution
-                             initializer=initializer)
+            _labels = labels.generate()
+            super().__init__(len(_labels), initializer=initializer)
+            self._register_labels(_labels)
 
-    def _initialize_labels(self, labels):
+    def _register_labels(self, labels):
-        if type(labels) == tuple:
+        self.register_buffer("_labels", labels)
-            num_classes, prototypes_per_class = labels
-            labels = EqualLabelInitializer(num_classes, prototypes_per_class)
 
-        self._labels = labels.generate()
+    @property
+    def distribution(self):
+        clabels, counts = torch.unique(self._labels,
+                                       sorted=True,
+                                       return_counts=True)
+        return dict(zip(clabels.tolist(), counts.tolist()))
+
+    def _initialize_components(self, num_components, initializer):
+        if isinstance(initializer, ClassAwareInitializer):
+            _precheck_initializer(initializer)
+            _components = initializer.generate(num_components,
+                                               self.initial_distribution)
+            self._register_components(_components)
+        else:
+            super()._initialize_components(num_components, initializer)
+
+    def add_components(self, distribution, initializer):
+        _precheck_initializer(initializer)
+
+        # Labels
+        labels = get_labels_object(distribution)
+        new_labels = labels.generate()
+        _labels = torch.cat([self._labels, new_labels])
+        self._register_labels(_labels)
+
+        # Components
+        if isinstance(initializer, ClassAwareInitializer):
+            _new = initializer.generate(len(new_labels), labels.distribution)
+        else:
+            _new = initializer.generate(len(new_labels))
+        _components = torch.cat([self._components, _new])
+        self._register_components(_components)
+
+    def remove_components(self, indices=None):
+        # Components
+        mask = super().remove_components(indices)
+
+        # Labels
+        _labels = self._labels[mask]
+        self._register_labels(_labels)
 
     @property
     def component_labels(self):
         """Tensor containing the component tensors."""
-        return self._labels.detach().cpu()
+        return self._labels.detach()
 
     def forward(self):
         return super().forward(), self._labels
@@ -106,20 +200,21 @@ class ReasoningComponents(Components):
                  *,
                  initialized_components=None):
         if initialized_components is not None:
-            super().__init__(initialized_components=initialized_components[0])
+            components, reasonings = initialized_components
-            self._reasonings = initialized_components[1]
+
+            super().__init__(initialized_components=components)
+            self.register_parameter("_reasonings", reasonings)
         else:
             self._initialize_reasonings(reasonings)
-            super().__init__(number_of_components=len(self._reasonings),
+            super().__init__(len(self._reasonings), initializer=initializer)
-                             initializer=initializer)
 
     def _initialize_reasonings(self, reasonings):
-        if type(reasonings) == tuple:
+        if isinstance(reasonings, tuple):
-            num_classes, number_of_components = reasonings
+            num_classes, num_components = reasonings
-            reasonings = ZeroReasoningsInitializer(num_classes,
+            reasonings = ZeroReasoningsInitializer(num_classes, num_components)
-                                                   number_of_components)
 
-        self._reasonings = reasonings.generate()
+        _reasonings = reasonings.generate()
+        self.register_parameter("_reasonings", _reasonings)
 
     @property
     def reasonings(self):
@@ -128,7 +223,7 @@ class ReasoningComponents(Components):
         Dimension NxCx2
 
         """
-        return self._reasonings.detach().cpu()
+        return self._reasonings.detach()
 
     def forward(self):
         return super().forward(), self._reasonings

prototorch/components/initializers.py

@@ -1,26 +1,43 @@
-"""ProtoTroch Initializers."""
+"""ProtoTroch Component and Label Initializers."""
+
 import warnings
 from collections.abc import Iterable
+from itertools import chain
 
 import torch
 from torch.utils.data import DataLoader, Dataset
 
 
-def parse_init_arg(arg):
+def parse_data_arg(data_arg):
-    if isinstance(arg, Dataset):
+    if isinstance(data_arg, Dataset):
-        data, labels = next(iter(DataLoader(arg, batch_size=len(arg))))
+        data_arg = DataLoader(data_arg, batch_size=len(data_arg))
-        # data = data.view(len(arg), -1)  # flatten
+
+    if isinstance(data_arg, DataLoader):
+        data = torch.tensor([])
+        targets = torch.tensor([])
+        for x, y in data_arg:
+            data = torch.cat([data, x])
+            targets = torch.cat([targets, y])
     else:
-        data, labels = arg
+        data, targets = data_arg
         if not isinstance(data, torch.Tensor):
             wmsg = f"Converting data to {torch.Tensor}."
             warnings.warn(wmsg)
             data = torch.Tensor(data)
-        if not isinstance(labels, torch.Tensor):
+        if not isinstance(targets, torch.Tensor):
-            wmsg = f"Converting labels to {torch.Tensor}."
+            wmsg = f"Converting targets to {torch.Tensor}."
             warnings.warn(wmsg)
-            labels = torch.Tensor(labels)
+            targets = torch.Tensor(targets)
-    return data, labels
+    return data, targets
+
+
+def get_subinitializers(data, targets, clabels, subinit_type):
+    initializers = dict()
+    for clabel in clabels:
+        class_data = data[targets == clabel]
+        class_initializer = subinit_type(class_data)
+        initializers[clabel] = (class_initializer)
+    return initializers
 
 
 # Components
@@ -30,18 +47,22 @@ class ComponentsInitializer(object):
 
 
 class DimensionAwareInitializer(ComponentsInitializer):
-    def __init__(self, c_dims):
+    def __init__(self, dims):
         super().__init__()
-        if isinstance(c_dims, Iterable):
+        if isinstance(dims, Iterable):
-            self.components_dims = tuple(c_dims)
+            self.components_dims = tuple(dims)
         else:
-            self.components_dims = (c_dims, )
+            self.components_dims = (dims, )
 
 
 class OnesInitializer(DimensionAwareInitializer):
+    def __init__(self, dims, scale=1.0):
+        super().__init__(dims)
+        self.scale = scale
+
     def generate(self, length):
         gen_dims = (length, ) + self.components_dims
-        return torch.ones(gen_dims)
+        return torch.ones(gen_dims) * self.scale
 
 
 class ZerosInitializer(DimensionAwareInitializer):
@@ -51,88 +72,99 @@ class ZerosInitializer(DimensionAwareInitializer):
 
 
 class UniformInitializer(DimensionAwareInitializer):
-    def __init__(self, c_dims, min=0.0, max=1.0):
+    def __init__(self, dims, minimum=0.0, maximum=1.0, scale=1.0):
-        super().__init__(c_dims)
+        super().__init__(dims)
-
+        self.minimum = minimum
-        self.min = min
+        self.maximum = maximum
-        self.max = max
+        self.scale = scale
 
     def generate(self, length):
         gen_dims = (length, ) + self.components_dims
-        return torch.ones(gen_dims).uniform_(self.min, self.max)
+        return torch.ones(gen_dims).uniform_(self.minimum,
+                                             self.maximum) * self.scale
 
 
-class PositionAwareInitializer(ComponentsInitializer):
+class DataAwareInitializer(ComponentsInitializer):
-    def __init__(self, positions):
+    def __init__(self, data, transform=torch.nn.Identity()):
         super().__init__()
-        self.data = positions
+        self.data = data
+        self.transform = transform
+
+    def __del__(self):
+        del self.data
 
 
-class SelectionInitializer(PositionAwareInitializer):
+class SelectionInitializer(DataAwareInitializer):
     def generate(self, length):
         indices = torch.LongTensor(length).random_(0, len(self.data))
-        return self.data[indices]
+        return self.transform(self.data[indices])
 
 
-class MeanInitializer(PositionAwareInitializer):
+class MeanInitializer(DataAwareInitializer):
     def generate(self, length):
         mean = torch.mean(self.data, dim=0)
         repeat_dim = [length] + [1] * len(mean.shape)
-        return mean.repeat(repeat_dim)
+        return self.transform(mean.repeat(repeat_dim))
 
 
-class ClassAwareInitializer(ComponentsInitializer):
+class ClassAwareInitializer(DataAwareInitializer):
-    def __init__(self, arg):
+    def __init__(self, data, transform=torch.nn.Identity()):
-        super().__init__()
+        data, targets = parse_data_arg(data)
-        data, labels = parse_init_arg(arg)
+        super().__init__(data, transform)
-        self.data = data
+        self.targets = targets
-        self.labels = labels
+        self.clabels = torch.unique(self.targets).int().tolist()
-
-        self.clabels = torch.unique(self.labels)
         self.num_classes = len(self.clabels)
 
+    def _get_samples_from_initializer(self, length, dist):
+        if not dist:
+            per_class = length // self.num_classes
+            dist = dict(zip(self.clabels, self.num_classes * [per_class]))
+        if isinstance(dist, list):
+            dist = dict(zip(self.clabels, dist))
+        samples = [self.initializers[k].generate(n) for k, n in dist.items()]
+        out = torch.vstack(samples)
+        with torch.no_grad():
+            out = self.transform(out)
+        return out
+
+    def __del__(self):
+        del self.data
+        del self.targets
+
 
 class StratifiedMeanInitializer(ClassAwareInitializer):
-    def __init__(self, arg):
+    def __init__(self, data, **kwargs):
-        super().__init__(arg)
+        super().__init__(data, **kwargs)
+        self.initializers = get_subinitializers(self.data, self.targets,
+                                                self.clabels, MeanInitializer)
 
-        self.initializers = []
+    def generate(self, length, dist):
-        for clabel in self.clabels:
+        samples = self._get_samples_from_initializer(length, dist)
-            class_data = self.data[self.labels == clabel]
+        return samples
-            class_initializer = MeanInitializer(class_data)
-            self.initializers.append(class_initializer)
-
-    def generate(self, length):
-        per_class = length // self.num_classes
-        samples_list = [init.generate(per_class) for init in self.initializers]
-        return torch.vstack(samples_list)
 
 
 class StratifiedSelectionInitializer(ClassAwareInitializer):
-    def __init__(self, arg, *, noise=None):
+    def __init__(self, data, noise=None, **kwargs):
-        super().__init__(arg)
+        super().__init__(data, **kwargs)
         self.noise = noise
+        self.initializers = get_subinitializers(self.data, self.targets,
+                                                self.clabels,
+                                                SelectionInitializer)
 
-        self.initializers = []
+    def add_noise_v1(self, x):
-        for clabel in self.clabels:
+        return x + self.noise
-            class_data = self.data[self.labels == clabel]
-            class_initializer = SelectionInitializer(class_data)
-            self.initializers.append(class_initializer)
 
-    def add_noise(self, x):
+    def add_noise_v2(self, x):
         """Shifts some dimensions of the data randomly."""
         n1 = torch.rand_like(x)
         n2 = torch.rand_like(x)
         mask = torch.bernoulli(n1) - torch.bernoulli(n2)
         return x + (self.noise * mask)
 
-    def generate(self, length):
+    def generate(self, length, dist):
-        per_class = length // self.num_classes
+        samples = self._get_samples_from_initializer(length, dist)
-        samples_list = [init.generate(per_class) for init in self.initializers]
-        samples = torch.vstack(samples_list)
         if self.noise is not None:
-            # samples = self.add_noise(samples)
+            samples = self.add_noise_v1(samples)
-            samples = samples + self.noise
         return samples
 
 
@@ -142,15 +174,43 @@ class LabelsInitializer:
         raise NotImplementedError("Subclasses should implement this!")
 
 
-class EqualLabelInitializer(LabelsInitializer):
+class UnequalLabelsInitializer(LabelsInitializer):
+    def __init__(self, dist):
+        self.dist = dist
+
+    @property
+    def distribution(self):
+        return self.dist
+
+    def generate(self, clabels=None, dist=None):
+        if not clabels:
+            clabels = range(len(self.dist))
+        if not dist:
+            dist = self.dist
+        targets = list(chain(*[[i] * n for i, n in zip(clabels, dist)]))
+        return torch.LongTensor(targets)
+
+
+class EqualLabelsInitializer(LabelsInitializer):
     def __init__(self, classes, per_class):
         self.classes = classes
         self.per_class = per_class
 
+    @property
+    def distribution(self):
+        return self.classes * [self.per_class]
+
     def generate(self):
         return torch.arange(self.classes).repeat(self.per_class, 1).T.flatten()
 
 
+class CustomLabelsInitializer(UnequalLabelsInitializer):
+    def generate(self):
+        clabels = list(self.dist.keys())
+        dist = list(self.dist.values())
+        return super().generate(clabels, dist)
+
+
 # Reasonings
 class ReasoningsInitializer:
     def generate(self, length):
@@ -170,3 +230,5 @@ class ZeroReasoningsInitializer(ReasoningsInitializer):
 SSI = StratifiedSampleInitializer = StratifiedSelectionInitializer
 SMI = StratifiedMeanInitializer
 Random = RandomInitializer = UniformInitializer
+Zeros = ZerosInitializer
+Ones = OnesInitializer

prototorch/datasets/__init__.py

@@ -1,11 +1,6 @@
 """ProtoTorch datasets."""
 
 from .abstract import NumpyDataset
+from .sklearn import Blobs, Circles, Iris, Moons, Random
 from .spiral import Spiral
 from .tecator import Tecator
-
-__all__ = [
-    "NumpyDataset",
-    "Spiral",
-    "Tecator",
-]

prototorch/datasets/abstract.py

@@ -14,8 +14,10 @@ import torch
 
 class NumpyDataset(torch.utils.data.TensorDataset):
     """Create a PyTorch TensorDataset from NumPy arrays."""
-    def __init__(self, *arrays):
+    def __init__(self, data, targets):
-        tensors = [torch.Tensor(arr) for arr in arrays]
+        self.data = torch.Tensor(data)
+        self.targets = torch.LongTensor(targets)
+        tensors = [self.data, self.targets]
         super().__init__(*tensors)
 
 

prototorch/datasets/sklearn.py

@@ -0,0 +1,137 @@
+"""Thin wrappers for a few scikit-learn datasets.
+
+URL:
+    https://scikit-learn.org/stable/modules/classes.html#module-sklearn.datasets
+
+"""
+
+import warnings
+from typing import Sequence, Union
+
+from prototorch.datasets.abstract import NumpyDataset
+
+from sklearn.datasets import (load_iris, make_blobs, make_circles,
+                              make_classification, make_moons)
+
+
+class Iris(NumpyDataset):
+    """Iris Dataset by Ronald Fisher introduced in 1936.
+
+    The dataset contains four measurements from flowers of three species of iris.
+
+    .. list-table:: Iris
+        :header-rows: 1
+
+        * - dimensions
+          - classes
+          - training size
+          - validation size
+          - test size
+        * - 4
+          - 3
+          - 150
+          - 0
+          - 0
+
+    :param dims: select a subset of dimensions
+    """
+    def __init__(self, dims: Sequence[int] = None):
+        x, y = load_iris(return_X_y=True)
+        if dims:
+            x = x[:, dims]
+        super().__init__(x, y)
+
+
+class Blobs(NumpyDataset):
+    """Generate isotropic Gaussian blobs for clustering.
+
+    Read more at
+    https://scikit-learn.org/stable/datasets/sample_generators.html#sample-generators.
+
+    """
+    def __init__(self,
+                 num_samples: int = 300,
+                 num_features: int = 2,
+                 seed: Union[None, int] = 0):
+        x, y = make_blobs(num_samples,
+                          num_features,
+                          centers=None,
+                          random_state=seed,
+                          shuffle=False)
+        super().__init__(x, y)
+
+
+class Random(NumpyDataset):
+    """Generate a random n-class classification problem.
+
+    Read more at
+    https://scikit-learn.org/stable/modules/generated/sklearn.datasets.make_classification.html.
+
+    Note: n_classes * n_clusters_per_class <= 2**n_informative must satisfy.
+    """
+    def __init__(self,
+                 num_samples: int = 300,
+                 num_features: int = 2,
+                 num_classes: int = 2,
+                 num_clusters: int = 2,
+                 num_informative: Union[None, int] = None,
+                 separation: float = 1.0,
+                 seed: Union[None, int] = 0):
+        if not num_informative:
+            import math
+            num_informative = math.ceil(math.log2(num_classes * num_clusters))
+            if num_features < num_informative:
+                warnings.warn("Generating more features than requested.")
+                num_features = num_informative
+        x, y = make_classification(num_samples,
+                                   num_features,
+                                   n_informative=num_informative,
+                                   n_redundant=0,
+                                   n_classes=num_classes,
+                                   n_clusters_per_class=num_clusters,
+                                   class_sep=separation,
+                                   random_state=seed,
+                                   shuffle=False)
+        super().__init__(x, y)
+
+
+class Circles(NumpyDataset):
+    """Make a large circle containing a smaller circle in 2D.
+
+    A simple toy dataset to visualize clustering and classification algorithms.
+
+    Read more at
+    https://scikit-learn.org/stable/modules/generated/sklearn.datasets.make_circles.html
+
+    """
+    def __init__(self,
+                 num_samples: int = 300,
+                 noise: float = 0.3,
+                 factor: float = 0.8,
+                 seed: Union[None, int] = 0):
+        x, y = make_circles(num_samples,
+                            noise=noise,
+                            factor=factor,
+                            random_state=seed,
+                            shuffle=False)
+        super().__init__(x, y)
+
+
+class Moons(NumpyDataset):
+    """Make two interleaving half circles.
+
+    A simple toy dataset to visualize clustering and classification algorithms.
+
+    Read more at
+    https://scikit-learn.org/stable/modules/generated/sklearn.datasets.make_moons.html
+
+    """
+    def __init__(self,
+                 num_samples: int = 300,
+                 noise: float = 0.3,
+                 seed: Union[None, int] = 0):
+        x, y = make_moons(num_samples,
+                          noise=noise,
+                          random_state=seed,
+                          shuffle=False)
+        super().__init__(x, y)

prototorch/datasets/spiral.py

@@ -4,18 +4,22 @@ import numpy as np
 import torch
 
 
-def make_spiral(n_samples=500, noise=0.3):
+def make_spiral(num_samples=500, noise=0.3):
+    """Generates the Spiral Dataset.
+
+    For use in Prototorch use `prototorch.datasets.Spiral` instead.
+    """
     def get_samples(n, delta_t):
         points = []
         for i in range(n):
-            r = i / n_samples * 5
+            r = i / num_samples * 5
             t = 1.75 * i / n * 2 * np.pi + delta_t
             x = r * np.sin(t) + np.random.rand(1) * noise
             y = r * np.cos(t) + np.random.rand(1) * noise
             points.append([x, y])
         return points
 
-    n = n_samples // 2
+    n = num_samples // 2
     positive = get_samples(n=n, delta_t=0)
     negative = get_samples(n=n, delta_t=np.pi)
     x = np.concatenate(
@@ -27,7 +31,27 @@ def make_spiral(n_samples=500, noise=0.3):
 
 
 class Spiral(torch.utils.data.TensorDataset):
-    """Spiral dataset for binary classification."""
+    """Spiral dataset for binary classification.
-    def __init__(self, n_samples=500, noise=0.3):
+
-        x, y = make_spiral(n_samples, noise)
+    This datasets consists of two spirals of two different classes.
+
+    .. list-table:: Spiral
+        :header-rows: 1
+
+        * - dimensions
+          - classes
+          - training size
+          - validation size
+          - test size
+        * - 2
+          - 2
+          - num_samples
+          - 0
+          - 0
+
+    :param num_samples: number of random samples
+    :param noise: noise added to the spirals
+    """
+    def __init__(self, num_samples: int = 500, noise: float = 0.3):
+        x, y = make_spiral(num_samples, noise)
         super().__init__(torch.Tensor(x), torch.LongTensor(y))

prototorch/datasets/tecator.py

@@ -40,15 +40,29 @@ import os
 
 import numpy as np
 import torch
-from torchvision.datasets.utils import download_file_from_google_drive
-
 from prototorch.datasets.abstract import ProtoDataset
+from torchvision.datasets.utils import download_file_from_google_drive
 
 
 class Tecator(ProtoDataset):
     """
-    `Tecator Dataset <http://lib.stat.cmu.edu/datasets/tecator>`__
+    `Tecator Dataset <http://lib.stat.cmu.edu/datasets/tecator>`__ for classification.
-    for classification.
+
+    The dataset contains wavelength measurements of meat.
+
+    .. list-table:: Tecator
+        :header-rows: 1
+
+        * - dimensions
+          - classes
+          - training size
+          - validation size
+          - test size
+        * - 100
+          - 2
+          - 129
+          - 43
+          - 43
     """
 
     _resources = [
@@ -87,12 +101,12 @@ class Tecator(ProtoDataset):
             x_train, y_train = f["x_train"], f["y_train"]
             x_test, y_test = f["x_test"], f["y_test"]
         training_set = [
-            torch.tensor(x_train, dtype=torch.float32),
+            torch.Tensor(x_train),
-            torch.tensor(y_train),
+            torch.LongTensor(y_train),
         ]
         test_set = [
-            torch.tensor(x_test, dtype=torch.float32),
+            torch.Tensor(x_test),
-            torch.tensor(y_test),
+            torch.LongTensor(y_test),
         ]
 
         with open(os.path.join(self.processed_folder, self.training_file),

prototorch/functions/__init__.py

@@ -2,11 +2,4 @@
 
 from .activations import identity, sigmoid_beta, swish_beta
 from .competitions import knnc, wtac
-
+from .pooling import *
-__all__ = [
-    "identity",
-    "sigmoid_beta",
-    "swish_beta",
-    "knnc",
-    "wtac",
-]

prototorch/functions/activations.py

@@ -5,17 +5,14 @@ import torch
 ACTIVATIONS = dict()
 
 
-# def register_activation(scriptf):
+def register_activation(fn):
-#     ACTIVATIONS[scriptf.name] = scriptf
-#     return scriptf
-def register_activation(function):
     """Add the activation function to the registry."""
-    ACTIVATIONS[function.__name__] = function
+    name = fn.__name__
-    return function
+    ACTIVATIONS[name] = fn
+    return fn
 
 
 @register_activation
-# @torch.jit.script
 def identity(x, beta=0.0):
     """Identity activation function.
 
@@ -29,7 +26,6 @@ def identity(x, beta=0.0):
 
 
 @register_activation
-# @torch.jit.script
 def sigmoid_beta(x, beta=10.0):
     r"""Sigmoid activation function with scaling.
 
@@ -44,7 +40,6 @@ def sigmoid_beta(x, beta=10.0):
 
 
 @register_activation
-# @torch.jit.script
 def swish_beta(x, beta=10.0):
     r"""Swish activation function with scaling.
 

prototorch/functions/competitions.py

@@ -3,43 +3,26 @@
 import torch
 
 
-# @torch.jit.script
+def wtac(distances: torch.Tensor,
-def stratified_min(distances, labels):
+         labels: torch.LongTensor) -> (torch.LongTensor):
-    clabels = torch.unique(labels, dim=0)
+    """Winner-Takes-All-Competition.
-    nclasses = clabels.size()[0]
-    if distances.size()[1] == nclasses:
-        # skip if only one prototype per class
-        return distances
-    batch_size = distances.size()[0]
-    winning_distances = torch.zeros(nclasses, batch_size)
-    inf = torch.full_like(distances.T, fill_value=float("inf"))
-    # distances_to_wpluses = torch.where(matcher, distances, inf)
-    for i, cl in enumerate(clabels):
-        # cdists = distances.T[labels == cl]
-        matcher = torch.eq(labels.unsqueeze(dim=1), cl)
-        if labels.ndim == 2:
-            # if the labels are one-hot vectors
-            matcher = torch.eq(torch.sum(matcher, dim=-1), nclasses)
-        cdists = torch.where(matcher, distances.T, inf).T
-        winning_distances[i] = torch.min(cdists, dim=1,
-                                         keepdim=True).values.squeeze()
-    if labels.ndim == 2:
-        # Transpose to return with `batch_size` first and
-        # reverse the columns to fix the ordering of the classes
-        return torch.flip(winning_distances.T, dims=(1, ))
 
-    return winning_distances.T  # return with `batch_size` first
+    Returns the labels corresponding to the winners.
 
-
+    """
-# @torch.jit.script
-def wtac(distances, labels):
     winning_indices = torch.min(distances, dim=1).indices
     winning_labels = labels[winning_indices].squeeze()
     return winning_labels
 
 
-# @torch.jit.script
+def knnc(distances: torch.Tensor,
-def knnc(distances, labels, k):
+         labels: torch.LongTensor,
-    winning_indices = torch.topk(-distances, k=k.item(), dim=1).indices
+         k: int = 1) -> (torch.LongTensor):
-    winning_labels = labels[winning_indices].squeeze()
+    """K-Nearest-Neighbors-Competition.
+
+    Returns the labels corresponding to the winners.
+
+    """
+    winning_indices = torch.topk(-distances, k=k, dim=1).indices
+    winning_labels = torch.mode(labels[winning_indices], dim=1).values
     return winning_labels

prototorch/functions/distances.py

@@ -2,9 +2,8 @@
 
 import numpy as np
 import torch
-
 from prototorch.functions.helper import (_check_shapes, _int_and_mixed_shape,
-                                         equal_int_shape)
+                                         equal_int_shape, get_flat)
 
 
 def squared_euclidean_distance(x, y):
@@ -12,12 +11,10 @@ def squared_euclidean_distance(x, y):
 
     Compute :math:`{\langle \bm x - \bm y \rangle}_2`
 
-    :param `torch.tensor` x: Two dimensional vector
-    :param `torch.tensor` y: Two dimensional vector
-
     **Alias:**
     ``prototorch.functions.distances.sed``
     """
+    x, y = get_flat(x, y)
     expanded_x = x.unsqueeze(dim=1)
     batchwise_difference = y - expanded_x
     differences_raised = torch.pow(batchwise_difference, 2)
@@ -30,18 +27,17 @@ def euclidean_distance(x, y):
 
     Compute :math:`\sqrt{{\langle \bm x - \bm y \rangle}_2}`
 
-    :param `torch.tensor` x: Input Tensor of shape :math:`X \times N`
-    :param `torch.tensor` y: Input Tensor of shape :math:`Y \times N`
-
     :returns: Distance Tensor of shape :math:`X \times Y`
     :rtype: `torch.tensor`
     """
+    x, y = get_flat(x, y)
     distances_raised = squared_euclidean_distance(x, y)
     distances = torch.sqrt(distances_raised)
     return distances
 
 
 def euclidean_distance_v2(x, y):
+    x, y = get_flat(x, y)
     diff = y - x.unsqueeze(1)
     pairwise_distances = (diff @ diff.permute((0, 2, 1))).sqrt()
     # Passing `dim1=-2` and `dim2=-1` to `diagonal()` takes the
@@ -62,10 +58,9 @@ def lpnorm_distance(x, y, p):
 
     Calls ``torch.cdist``
 
-    :param `torch.tensor` x: Two dimensional vector
-    :param `torch.tensor` y: Two dimensional vector
     :param p: p parameter of the lp norm
     """
+    x, y = get_flat(x, y)
     distances = torch.cdist(x, y, p=p)
     return distances
 
@@ -75,10 +70,9 @@ def omega_distance(x, y, omega):
 
     Compute :math:`{\| \Omega \bm x - \Omega \bm y \|}_p`
 
-    :param `torch.tensor` x: Two dimensional vector
-    :param `torch.tensor` y: Two dimensional vector
     :param `torch.tensor` omega: Two dimensional matrix
     """
+    x, y = get_flat(x, y)
     projected_x = x @ omega
     projected_y = y @ omega
     distances = squared_euclidean_distance(projected_x, projected_y)
@@ -90,10 +84,9 @@ def lomega_distance(x, y, omegas):
 
     Compute :math:`{\| \Omega_k \bm x - \Omega_k \bm y_k \|}_p`
 
-    :param `torch.tensor` x: Two dimensional vector
-    :param `torch.tensor` y: Two dimensional vector
     :param `torch.tensor` omegas: Three dimensional matrix
     """
+    x, y = get_flat(x, y)
     projected_x = x @ omegas
     projected_y = torch.diagonal(y @ omegas).T
     expanded_y = torch.unsqueeze(projected_y, dim=1)

prototorch/functions/helper.py

@@ -1,6 +1,11 @@
 import torch
 
 
+def get_flat(*args):
+    rv = [x.view(x.size(0), -1) for x in args]
+    return rv
+
+
 def calculate_prototype_accuracy(y_pred, y_true, plabels):
     """Computes the accuracy of a prototype based model.
     via Winner-Takes-All rule.

prototorch/functions/initializers.py

@@ -15,59 +15,59 @@ def register_initializer(function):
 
 def labels_from(distribution, one_hot=True):
     """Takes a distribution tensor and returns a labels tensor."""
-    nclasses = distribution.shape[0]
+    num_classes = distribution.shape[0]
-    llist = [[i] * n for i, n in zip(range(nclasses), distribution)]
+    llist = [[i] * n for i, n in zip(range(num_classes), distribution)]
     # labels = [l for cl in llist for l in cl]  # flatten the list of lists
     flat_llist = list(chain(*llist))  # flatten label list with itertools.chain
     plabels = torch.tensor(flat_llist, requires_grad=False)
     if one_hot:
-        return torch.eye(nclasses)[plabels]
+        return torch.eye(num_classes)[plabels]
     return plabels
 
 
 @register_initializer
 def ones(x_train, y_train, prototype_distribution, one_hot=True):
-    nprotos = torch.sum(prototype_distribution)
+    num_protos = torch.sum(prototype_distribution)
-    protos = torch.ones(nprotos, *x_train.shape[1:])
+    protos = torch.ones(num_protos, *x_train.shape[1:])
     plabels = labels_from(prototype_distribution, one_hot)
     return protos, plabels
 
 
 @register_initializer
 def zeros(x_train, y_train, prototype_distribution, one_hot=True):
-    nprotos = torch.sum(prototype_distribution)
+    num_protos = torch.sum(prototype_distribution)
-    protos = torch.zeros(nprotos, *x_train.shape[1:])
+    protos = torch.zeros(num_protos, *x_train.shape[1:])
     plabels = labels_from(prototype_distribution, one_hot)
     return protos, plabels
 
 
 @register_initializer
 def rand(x_train, y_train, prototype_distribution, one_hot=True):
-    nprotos = torch.sum(prototype_distribution)
+    num_protos = torch.sum(prototype_distribution)
-    protos = torch.rand(nprotos, *x_train.shape[1:])
+    protos = torch.rand(num_protos, *x_train.shape[1:])
     plabels = labels_from(prototype_distribution, one_hot)
     return protos, plabels
 
 
 @register_initializer
 def randn(x_train, y_train, prototype_distribution, one_hot=True):
-    nprotos = torch.sum(prototype_distribution)
+    num_protos = torch.sum(prototype_distribution)
-    protos = torch.randn(nprotos, *x_train.shape[1:])
+    protos = torch.randn(num_protos, *x_train.shape[1:])
     plabels = labels_from(prototype_distribution, one_hot)
     return protos, plabels
 
 
 @register_initializer
 def stratified_mean(x_train, y_train, prototype_distribution, one_hot=True):
-    nprotos = torch.sum(prototype_distribution)
+    num_protos = torch.sum(prototype_distribution)
     pdim = x_train.shape[1]
-    protos = torch.empty(nprotos, pdim)
+    protos = torch.empty(num_protos, pdim)
     plabels = labels_from(prototype_distribution, one_hot)
     for i, label in enumerate(plabels):
         matcher = torch.eq(label.unsqueeze(dim=0), y_train)
         if one_hot:
-            nclasses = y_train.size()[1]
+            num_classes = y_train.size()[1]
-            matcher = torch.eq(torch.sum(matcher, dim=-1), nclasses)
+            matcher = torch.eq(torch.sum(matcher, dim=-1), num_classes)
         xl = x_train[matcher]
         mean_xl = torch.mean(xl, dim=0)
         protos[i] = mean_xl
@@ -81,15 +81,15 @@ def stratified_random(x_train,
                       prototype_distribution,
                       one_hot=True,
                       epsilon=1e-7):
-    nprotos = torch.sum(prototype_distribution)
+    num_protos = torch.sum(prototype_distribution)
     pdim = x_train.shape[1]
-    protos = torch.empty(nprotos, pdim)
+    protos = torch.empty(num_protos, pdim)
     plabels = labels_from(prototype_distribution, one_hot)
     for i, label in enumerate(plabels):
         matcher = torch.eq(label.unsqueeze(dim=0), y_train)
         if one_hot:
-            nclasses = y_train.size()[1]
+            num_classes = y_train.size()[1]
-            matcher = torch.eq(torch.sum(matcher, dim=-1), nclasses)
+            matcher = torch.eq(torch.sum(matcher, dim=-1), num_classes)
         xl = x_train[matcher]
         rand_index = torch.zeros(1).long().random_(0, xl.shape[0] - 1)
         random_xl = xl[rand_index]

prototorch/functions/losses.py

@@ -8,12 +8,12 @@ def _get_matcher(targets, labels):
     matcher = torch.eq(targets.unsqueeze(dim=1), labels)
     if labels.ndim == 2:
         # if the labels are one-hot vectors
-        nclasses = targets.size()[1]
+        num_classes = targets.size()[1]
-        matcher = torch.eq(torch.sum(matcher, dim=-1), nclasses)
+        matcher = torch.eq(torch.sum(matcher, dim=-1), num_classes)
     return matcher
 
 
-def _get_dp_dm(distances, targets, plabels):
+def _get_dp_dm(distances, targets, plabels, with_indices=False):
     """Returns the d+ and d- values for a batch of distances."""
     matcher = _get_matcher(targets, plabels)
     not_matcher = torch.bitwise_not(matcher)
@@ -21,9 +21,11 @@ def _get_dp_dm(distances, targets, plabels):
     inf = torch.full_like(distances, fill_value=float("inf"))
     d_matching = torch.where(matcher, distances, inf)
     d_unmatching = torch.where(not_matcher, distances, inf)
-    dp = torch.min(d_matching, dim=1, keepdim=True).values
+    dp = torch.min(d_matching, dim=-1, keepdim=True)
-    dm = torch.min(d_unmatching, dim=1, keepdim=True).values
+    dm = torch.min(d_unmatching, dim=-1, keepdim=True)
-    return dp, dm
+    if with_indices:
+        return dp, dm
+    return dp.values, dm.values
 
 
 def glvq_loss(distances, target_labels, prototype_labels):
@@ -47,10 +49,46 @@ def lvq1_loss(distances, target_labels, prototype_labels):
 
 def lvq21_loss(distances, target_labels, prototype_labels):
     """LVQ2.1 loss function with support for one-hot labels.
-    
+
     See Section 4 [Sado&Yamada]
     https://papers.nips.cc/paper/1995/file/9c3b1830513cc3b8fc4b76635d32e692-Paper.pdf
     """
     dp, dm = _get_dp_dm(distances, target_labels, prototype_labels)
     mu = dp - dm
-    return mu
+
+    return mu
+
+
+# Probabilistic
+def _get_class_probabilities(probabilities, targets, prototype_labels):
+    # Create Label Mapping
+    uniques = prototype_labels.unique(sorted=True).tolist()
+    key_val = {key: val for key, val in zip(uniques, range(len(uniques)))}
+
+    target_indices = torch.LongTensor(list(map(key_val.get, targets.tolist())))
+
+    whole = probabilities.sum(dim=1)
+    correct = probabilities[torch.arange(len(probabilities)), target_indices]
+    wrong = whole - correct
+
+    return whole, correct, wrong
+
+
+def nllr_loss(probabilities, targets, prototype_labels):
+    """Compute the Negative Log-Likelihood Ratio loss."""
+    _, correct, wrong = _get_class_probabilities(probabilities, targets,
+                                                 prototype_labels)
+
+    likelihood = correct / wrong
+    log_likelihood = torch.log(likelihood)
+    return -1.0 * log_likelihood
+
+
+def rslvq_loss(probabilities, targets, prototype_labels):
+    """Compute the Robust Soft Learning Vector Quantization (RSLVQ) loss."""
+    whole, correct, _ = _get_class_probabilities(probabilities, targets,
+                                                 prototype_labels)
+
+    likelihood = correct / whole
+    log_likelihood = torch.log(likelihood)
+    return -1.0 * log_likelihood

prototorch/functions/pooling.py

@@ -0,0 +1,80 @@
+"""ProtoTorch pooling functions."""
+
+from typing import Callable
+
+import torch
+
+
+def stratify_with(values: torch.Tensor,
+                  labels: torch.LongTensor,
+                  fn: Callable,
+                  fill_value: float = 0.0) -> (torch.Tensor):
+    """Apply an arbitrary stratification strategy on the columns on `values`.
+
+    The outputs correspond to sorted labels.
+    """
+    clabels = torch.unique(labels, dim=0, sorted=True)
+    num_classes = clabels.size()[0]
+    if values.size()[1] == num_classes:
+        # skip if stratification is trivial
+        return values
+    batch_size = values.size()[0]
+    winning_values = torch.zeros(num_classes, batch_size, device=labels.device)
+    filler = torch.full_like(values.T, fill_value=fill_value)
+    for i, cl in enumerate(clabels):
+        matcher = torch.eq(labels.unsqueeze(dim=1), cl)
+        if labels.ndim == 2:
+            # if the labels are one-hot vectors
+            matcher = torch.eq(torch.sum(matcher, dim=-1), num_classes)
+        cdists = torch.where(matcher, values.T, filler).T
+        winning_values[i] = fn(cdists)
+    if labels.ndim == 2:
+        # Transpose to return with `batch_size` first and
+        # reverse the columns to fix the ordering of the classes
+        return torch.flip(winning_values.T, dims=(1, ))
+
+    return winning_values.T  # return with `batch_size` first
+
+
+def stratified_sum_pooling(values: torch.Tensor,
+                           labels: torch.LongTensor) -> (torch.Tensor):
+    """Group-wise sum."""
+    winning_values = stratify_with(
+        values,
+        labels,
+        fn=lambda x: torch.sum(x, dim=1, keepdim=True).squeeze(),
+        fill_value=0.0)
+    return winning_values
+
+
+def stratified_min_pooling(values: torch.Tensor,
+                           labels: torch.LongTensor) -> (torch.Tensor):
+    """Group-wise minimum."""
+    winning_values = stratify_with(
+        values,
+        labels,
+        fn=lambda x: torch.min(x, dim=1, keepdim=True).values.squeeze(),
+        fill_value=float("inf"))
+    return winning_values
+
+
+def stratified_max_pooling(values: torch.Tensor,
+                           labels: torch.LongTensor) -> (torch.Tensor):
+    """Group-wise maximum."""
+    winning_values = stratify_with(
+        values,
+        labels,
+        fn=lambda x: torch.max(x, dim=1, keepdim=True).values.squeeze(),
+        fill_value=-1.0 * float("inf"))
+    return winning_values
+
+
+def stratified_prod_pooling(values: torch.Tensor,
+                            labels: torch.LongTensor) -> (torch.Tensor):
+    """Group-wise maximum."""
+    winning_values = stratify_with(
+        values,
+        labels,
+        fn=lambda x: torch.prod(x, dim=1, keepdim=True).squeeze(),
+        fill_value=1.0)
+    return winning_values

prototorch/functions/transforms.py

@@ -0,0 +1,5 @@
+import torch
+
+
+def gaussian(distance, variance):
+    return torch.exp(-(distance * distance) / (2 * variance))

prototorch/modules/__init__.py

@@ -1,7 +1,7 @@
 """ProtoTorch modules."""
 
-from .prototypes import Prototypes1D
+from .competitions import *
-
+from .initializers import *
-__all__ = [
+from .pooling import *
-    "Prototypes1D",
+from .transformations import *
-]
+from .wrappers import LambdaLayer, LossLayer

prototorch/modules/competitions.py

@@ -0,0 +1,41 @@
+"""ProtoTorch Competition Modules."""
+
+import torch
+from prototorch.functions.competitions import knnc, wtac
+
+
+class WTAC(torch.nn.Module):
+    """Winner-Takes-All-Competition Layer.
+
+    Thin wrapper over the `wtac` function.
+
+    """
+    def forward(self, distances, labels):
+        return wtac(distances, labels)
+
+
+class LTAC(torch.nn.Module):
+    """Loser-Takes-All-Competition Layer.
+
+    Thin wrapper over the `wtac` function.
+
+    """
+    def forward(self, probs, labels):
+        return wtac(-1.0 * probs, labels)
+
+
+class KNNC(torch.nn.Module):
+    """K-Nearest-Neighbors-Competition.
+
+    Thin wrapper over the `knnc` function.
+
+    """
+    def __init__(self, k=1, **kwargs):
+        super().__init__(**kwargs)
+        self.k = k
+
+    def forward(self, distances, labels):
+        return knnc(distances, labels, k=self.k)
+
+    def extra_repr(self):
+        return f"k: {self.k}"

prototorch/modules/initializers.py

@@ -0,0 +1,61 @@
+"""ProtoTroch Module Initializers."""
+
+import torch
+
+
+# Transformations
+class MatrixInitializer(object):
+    def __init__(self, *args, **kwargs):
+        ...
+
+    def generate(self, shape):
+        raise NotImplementedError("Subclasses should implement this!")
+
+
+class ZerosInitializer(MatrixInitializer):
+    def generate(self, shape):
+        return torch.zeros(shape)
+
+
+class OnesInitializer(MatrixInitializer):
+    def __init__(self, scale=1.0):
+        super().__init__()
+        self.scale = scale
+
+    def generate(self, shape):
+        return torch.ones(shape) * self.scale
+
+
+class UniformInitializer(MatrixInitializer):
+    def __init__(self, minimum=0.0, maximum=1.0, scale=1.0):
+        super().__init__()
+        self.minimum = minimum
+        self.maximum = maximum
+        self.scale = scale
+
+    def generate(self, shape):
+        return torch.ones(shape).uniform_(self.minimum,
+                                          self.maximum) * self.scale
+
+
+class DataAwareInitializer(MatrixInitializer):
+    def __init__(self, data, transform=torch.nn.Identity()):
+        super().__init__()
+        self.data = data
+        self.transform = transform
+
+    def __del__(self):
+        del self.data
+
+
+class EigenVectorInitializer(DataAwareInitializer):
+    def generate(self, shape):
+        # TODO
+        raise NotImplementedError()
+
+
+# Aliases
+EV = EigenVectorInitializer
+Random = RandomInitializer = UniformInitializer
+Zeros = ZerosInitializer
+Ones = OnesInitializer

prototorch/modules/losses.py

@@ -1,7 +1,6 @@
 """ProtoTorch losses."""
 
 import torch
-
 from prototorch.functions.activations import get_activation
 from prototorch.functions.losses import glvq_loss
 
@@ -21,8 +20,8 @@ class GLVQLoss(torch.nn.Module):
 
 
 class NeuralGasEnergy(torch.nn.Module):
-    def __init__(self, lm):
+    def __init__(self, lm, **kwargs):
-        super().__init__()
+        super().__init__(**kwargs)
         self.lm = lm
 
     def forward(self, d):
@@ -38,3 +37,22 @@ class NeuralGasEnergy(torch.nn.Module):
     @staticmethod
     def _nghood_fn(rankings, lm):
         return torch.exp(-rankings / lm)
+
+
+class GrowingNeuralGasEnergy(NeuralGasEnergy):
+    def __init__(self, topology_layer, **kwargs):
+        super().__init__(**kwargs)
+        self.topology_layer = topology_layer
+
+    @staticmethod
+    def _nghood_fn(rankings, topology):
+        winner = rankings[:, 0]
+
+        weights = torch.zeros_like(rankings, dtype=torch.float)
+        weights[torch.arange(rankings.shape[0]), winner] = 1.0
+
+        neighbours = topology.get_neighbours(winner)
+
+        weights[neighbours] = 0.1
+
+        return weights

prototorch/modules/models.py

@@ -1,11 +1,8 @@
 import torch
-from torch import nn
+from prototorch.components import LabeledComponents, StratifiedMeanInitializer
-
+from prototorch.functions.distances import euclidean_distance_matrix
-from prototorch.functions.distances import (euclidean_distance_matrix,
-                                            tangent_distance)
-from prototorch.functions.helper import _check_shapes, _int_and_mixed_shape
 from prototorch.functions.normalization import orthogonalization
-from prototorch.modules.prototypes import Prototypes1D
+from torch import nn
 
 
 class GTLVQ(nn.Module):
@@ -80,45 +77,35 @@ class GTLVQ(nn.Module):
         super(GTLVQ, self).__init__()
 
         self.num_protos = num_classes * prototypes_per_class
+        self.num_protos_class = prototypes_per_class
         self.subspace_size = feature_dim if subspace_size is None else subspace_size
         self.feature_dim = feature_dim
+        self.num_classes = num_classes
+
+        cls_initializer = StratifiedMeanInitializer(prototype_data)
+        cls_distribution = {
+            "num_classes": num_classes,
+            "prototypes_per_class": prototypes_per_class,
+        }
+
+        self.cls = LabeledComponents(cls_distribution, cls_initializer)
 
         if subspace_data is None:
             raise ValueError("Init Data must be specified!")
 
         self.tpt = tangent_projection_type
         with torch.no_grad():
-            if self.tpt == "local" or self.tpt == "local_proj":
+            if self.tpt == "local":
-                self.init_local_subspace(subspace_data)
+                self.init_local_subspace(subspace_data, subspace_size,
+                                         self.num_protos)
             elif self.tpt == "global":
                 self.init_gobal_subspace(subspace_data, subspace_size)
             else:
                 self.subspaces = None
 
-        # Hypothesis-Margin-Classifier
-        self.cls = Prototypes1D(
-            input_dim=feature_dim,
-            prototypes_per_class=prototypes_per_class,
-            nclasses=num_classes,
-            prototype_initializer="stratified_mean",
-            data=prototype_data,
-        )
-
     def forward(self, x):
-        # Tangent Projection
+        if self.tpt == "local":
-        if self.tpt == "local_proj":
+            dis = self.local_tangent_distances(x)
-            x_conform = (x.unsqueeze(1).repeat_interleave(self.num_protos,
-                                                          1).unsqueeze(2))
-            dis, proj_x = self.local_tangent_projection(x_conform)
-
-            proj_x = proj_x.reshape(x.shape[0] * self.num_protos,
-                                    self.feature_dim)
-            return proj_x, dis
-        elif self.tpt == "local":
-            x_conform = (x.unsqueeze(1).repeat_interleave(self.num_protos,
-                                                          1).unsqueeze(2))
-            dis = tangent_distance(x_conform, self.cls.prototypes,
-                                   self.subspaces)
         elif self.tpt == "gloabl":
             dis = self.global_tangent_distances(x)
         else:
@@ -131,16 +118,14 @@ class GTLVQ(nn.Module):
         _, _, v = torch.svd(data)
         subspace = (torch.eye(v.shape[0]) - (v @ v.T)).T
         subspaces = subspace[:, :num_subspaces]
-        self.subspaces = (torch.nn.Parameter(
+        self.subspaces = nn.Parameter(subspaces, requires_grad=True)
-            subspaces).clone().detach().requires_grad_(True))
 
-    def init_local_subspace(self, data):
+    def init_local_subspace(self, data, num_subspaces, num_protos):
-        _, _, v = torch.svd(data)
+        data = data - torch.mean(data, dim=0)
-        inital_projector = (torch.eye(v.shape[0]) - (v @ v.T)).T
+        _, _, v = torch.svd(data, some=False)
-        subspaces = inital_projector.unsqueeze(0).repeat_interleave(
+        v = v[:, :num_subspaces]
-            self.num_protos, 0)
+        subspaces = v.unsqueeze(0).repeat_interleave(num_protos, 0)
-        self.subspaces = (torch.nn.Parameter(
+        self.subspaces = nn.Parameter(subspaces, requires_grad=True)
-            subspaces).clone().detach().requires_grad_(True))
 
     def global_tangent_distances(self, x):
         # Tangent Projection
@@ -151,37 +136,26 @@ class GTLVQ(nn.Module):
         # Euclidean Distance
         return euclidean_distance_matrix(x, projected_prototypes)
 
-    def local_tangent_projection(self, signals):
+    def local_tangent_distances(self, x):
-        # Note: subspaces is always assumed as transposed and must be orthogonal!
-        # shape(signals): batch x proto_number x channels x dim1 x dim2 x ... x dimN
-        # shape(protos): proto_number x dim1 x dim2 x ... x dimN
-        # shape(subspaces): (optional [proto_number]) x prod(dim1 * dim2 * ... * dimN)  x prod(projected_atom_shape)
-        # subspace should be orthogonalized
-        # Origin Source Code
-        # Origin Author:
-        protos = self.cls.prototypes
-        subspaces = self.subspaces
-        signal_shape, signal_int_shape = _int_and_mixed_shape(signals)
-        _, proto_int_shape = _int_and_mixed_shape(protos)
 
-        # check if the shapes are correct
+        # Tangent Distance
-        _check_shapes(signal_int_shape, proto_int_shape)
+        x = x.unsqueeze(1).expand(x.size(0), self.cls.num_components,
-
+                                  x.size(-1))
-        # Tangent Data Projections
+        protos = self.cls()[0].unsqueeze(0).expand(x.size(0),
-        projected_protos = torch.bmm(protos.unsqueeze(1), subspaces).squeeze(1)
+                                                   self.cls.num_components,
-        data = signals.squeeze(2).permute([1, 0, 2])
+                                                   x.size(-1))
-        projected_data = torch.bmm(data, subspaces)
+        projectors = torch.eye(
-        projected_data = projected_data.permute([1, 0, 2]).unsqueeze(1)
+            self.subspaces.shape[-2], device=x.device) - torch.bmm(
-        diff = projected_data - projected_protos
+                self.subspaces, self.subspaces.permute([0, 2, 1]))
-        projected_diff = torch.reshape(
+        diff = (x - protos)
-            diff, (signal_shape[1], signal_shape[0], signal_shape[2]) +
+        diff = diff.permute([1, 0, 2])
-            signal_shape[3:])
+        diff = torch.bmm(diff, projectors)
-        diss = torch.norm(projected_diff, 2, dim=-1)
+        diff = torch.norm(diff, 2, dim=-1).T
-        return diss.permute([1, 0, 2]).squeeze(-1), projected_data.squeeze(1)
+        return diff
 
     def get_parameters(self):
         return {
-            "params": self.cls.prototypes,
+            "params": self.cls.components,
         }, {
             "params": self.subspaces
         }

prototorch/modules/pooling.py

@@ -0,0 +1,31 @@
+"""ProtoTorch Pooling Modules."""
+
+import torch
+from prototorch.functions.pooling import (stratified_max_pooling,
+                                          stratified_min_pooling,
+                                          stratified_prod_pooling,
+                                          stratified_sum_pooling)
+
+
+class StratifiedSumPooling(torch.nn.Module):
+    """Thin wrapper over the `stratified_sum_pooling` function."""
+    def forward(self, values, labels):
+        return stratified_sum_pooling(values, labels)
+
+
+class StratifiedProdPooling(torch.nn.Module):
+    """Thin wrapper over the `stratified_prod_pooling` function."""
+    def forward(self, values, labels):
+        return stratified_prod_pooling(values, labels)
+
+
+class StratifiedMinPooling(torch.nn.Module):
+    """Thin wrapper over the `stratified_min_pooling` function."""
+    def forward(self, values, labels):
+        return stratified_min_pooling(values, labels)
+
+
+class StratifiedMaxPooling(torch.nn.Module):
+    """Thin wrapper over the `stratified_max_pooling` function."""
+    def forward(self, values, labels):
+        return stratified_max_pooling(values, labels)

prototorch/modules/prototypes.py

@@ -1,137 +0,0 @@
-"""ProtoTorch prototype modules."""
-
-import warnings
-
-import torch
-
-from prototorch.functions.initializers import get_initializer
-
-
-class _Prototypes(torch.nn.Module):
-    """Abstract prototypes class."""
-    def __init__(self, *args, **kwargs):
-        super().__init__(*args, **kwargs)
-
-    def _validate_prototype_distribution(self):
-        if 0 in self.prototype_distribution:
-            warnings.warn("Are you sure about the `0` in "
-                          "`prototype_distribution`?")
-
-    def extra_repr(self):
-        return f"prototypes.shape: {tuple(self.prototypes.shape)}"
-
-    def forward(self):
-        return self.prototypes, self.prototype_labels
-
-
-class Prototypes1D(_Prototypes):
-    """Create a learnable set of one-dimensional prototypes.
-
-    TODO Complete this doc-string.
-    """
-    def __init__(
-        self,
-        prototypes_per_class=1,
-        prototype_initializer="ones",
-        prototype_distribution=None,
-        data=None,
-        dtype=torch.float32,
-        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:
-            if not isinstance(prototype_distribution, list):
-                prototype_distribution = prototype_distribution.tolist()
-
-        if data is None:
-            if "input_dim" not in kwargs:
-                raise NameError("`input_dim` required if "
-                                "no `data` is provided.")
-            if prototype_distribution:
-                kwargs_nclasses = sum(prototype_distribution)
-            else:
-                if "nclasses" not in kwargs:
-                    raise NameError("`prototype_distribution` required if "
-                                    "both `data` and `nclasses` are not "
-                                    "provided.")
-                kwargs_nclasses = kwargs.pop("nclasses")
-            input_dim = kwargs.pop("input_dim")
-            if prototype_initializer in [
-                    "stratified_mean", "stratified_random"
-            ]:
-                warnings.warn(
-                    f"`prototype_initializer`: `{prototype_initializer}` "
-                    "requires `data`, but `data` is not provided. "
-                    "Using randomly generated data instead.")
-            x_train = torch.rand(kwargs_nclasses, input_dim)
-            y_train = torch.arange(kwargs_nclasses)
-            if one_hot_labels:
-                y_train = torch.eye(kwargs_nclasses)[y_train]
-            data = [x_train, y_train]
-
-        x_train, y_train = data
-        x_train = torch.as_tensor(x_train).type(dtype)
-        y_train = torch.as_tensor(y_train).type(torch.int)
-        nclasses = torch.unique(y_train, dim=-1).shape[-1]
-
-        if nclasses == 1:
-            warnings.warn("Are you sure about having one class only?")
-
-        if x_train.ndim != 2:
-            raise ValueError("`data[0].ndim != 2`.")
-
-        if y_train.ndim == 2:
-            if y_train.shape[1] == 1 and one_hot_labels:
-                raise ValueError("`one_hot_labels` is set to `True` "
-                                 "but target labels are not one-hot-encoded.")
-            if y_train.shape[1] != 1 and not one_hot_labels:
-                raise ValueError("`one_hot_labels` is set to `False` "
-                                 "but target labels in `data` "
-                                 "are one-hot-encoded.")
-        if y_train.ndim == 1 and one_hot_labels:
-            raise ValueError("`one_hot_labels` is set to `True` "
-                             "but target labels are not one-hot-encoded.")
-
-        # Verify input dimension if `input_dim` is provided
-        if "input_dim" in kwargs:
-            input_dim = kwargs.pop("input_dim")
-            if input_dim != x_train.shape[1]:
-                raise ValueError(f"Provided `input_dim`={input_dim} does "
-                                 "not match data dimension "
-                                 f"`data[0].shape[1]`={x_train.shape[1]}")
-
-        # Verify the number of classes if `nclasses` is provided
-        if "nclasses" in kwargs:
-            kwargs_nclasses = kwargs.pop("nclasses")
-            if kwargs_nclasses != nclasses:
-                raise ValueError(f"Provided `nclasses={kwargs_nclasses}` does "
-                                 "not match data labels "
-                                 "`torch.unique(data[1]).shape[0]`"
-                                 f"={nclasses}")
-
-        super().__init__(**kwargs)
-
-        if not prototype_distribution:
-            prototype_distribution = [prototypes_per_class] * nclasses
-        with torch.no_grad():
-            self.prototype_distribution = torch.tensor(prototype_distribution)
-
-        self._validate_prototype_distribution()
-
-        self.prototype_initializer = get_initializer(prototype_initializer)
-        prototypes, prototype_labels = self.prototype_initializer(
-            x_train,
-            y_train,
-            prototype_distribution=self.prototype_distribution,
-            one_hot=one_hot_labels,
-        )
-
-        # Register module parameters
-        self.prototypes = torch.nn.Parameter(prototypes)
-        self.prototype_labels = torch.nn.Parameter(
-            prototype_labels.type(dtype)).requires_grad_(False)

prototorch/modules/transformations.py

@@ -0,0 +1,49 @@
+"""ProtoTorch Transformation Layers."""
+
+import torch
+from torch.nn.parameter import Parameter
+
+from .initializers import MatrixInitializer
+
+
+def _precheck_initializer(initializer):
+    if not isinstance(initializer, MatrixInitializer):
+        emsg = f"`initializer` has to be some subtype of " \
+            f"{MatrixInitializer}. " \
+            f"You have provided: {initializer=} instead."
+        raise TypeError(emsg)
+
+
+class Omega(torch.nn.Module):
+    """The Omega mapping used in GMLVQ."""
+    def __init__(self,
+                 num_replicas=1,
+                 input_dim=None,
+                 latent_dim=None,
+                 initializer=None,
+                 *,
+                 initialized_weights=None):
+        super().__init__()
+
+        if initialized_weights is not None:
+            self._register_weights(initialized_weights)
+        else:
+            if num_replicas == 1:
+                shape = (input_dim, latent_dim)
+            else:
+                shape = (num_replicas, input_dim, latent_dim)
+            self._initialize_weights(shape, initializer)
+
+    def _register_weights(self, weights):
+        self.register_parameter("_omega", Parameter(weights))
+
+    def _initialize_weights(self, shape, initializer):
+        _precheck_initializer(initializer)
+        _omega = initializer.generate(shape)
+        self._register_weights(_omega)
+
+    def forward(self):
+        return self._omega
+
+    def extra_repr(self):
+        return f"(omega): (shape: {tuple(self._omega.shape)})"

prototorch/modules/wrappers.py

@@ -0,0 +1,36 @@
+"""ProtoTorch Wrappers."""
+
+import torch
+
+
+class LambdaLayer(torch.nn.Module):
+    def __init__(self, fn, name=None):
+        super().__init__()
+        self.fn = fn
+        self.name = name or fn.__name__  # lambda fns get <lambda>
+
+    def forward(self, *args, **kwargs):
+        return self.fn(*args, **kwargs)
+
+    def extra_repr(self):
+        return self.name
+
+
+class LossLayer(torch.nn.modules.loss._Loss):
+    def __init__(self,
+                 fn,
+                 name=None,
+                 size_average=None,
+                 reduce=None,
+                 reduction: str = "mean") -> None:
+        super().__init__(size_average=size_average,
+                         reduce=reduce,
+                         reduction=reduction)
+        self.fn = fn
+        self.name = name or fn.__name__  # lambda fns get <lambda>
+
+    def forward(self, *args, **kwargs):
+        return self.fn(*args, **kwargs)
+
+    def extra_repr(self):
+        return self.name

prototorch/utils/utils.py

@@ -1,243 +0,0 @@
-"""Utilities that provide various small functionalities."""
-
-import os
-import pickle
-import sys
-from time import time
-
-import matplotlib.pyplot as plt
-import numpy as np
-
-
-def progressbar(title, value, end, bar_width=20):
-    percent = float(value) / end
-    arrow = "=" * int(round(percent * bar_width) - 1) + ">"
-    spaces = "." * (bar_width - len(arrow))
-    sys.stdout.write("\r{}: [{}] {}%".format(title, arrow + spaces,
-                                             int(round(percent * 100))))
-    sys.stdout.flush()
-    if percent == 1.0:
-        print()
-
-
-def prettify_string(inputs, start="", sep=" ", end="\n"):
-    outputs = start + " ".join(inputs.split()) + end
-    return outputs
-
-
-def pretty_print(inputs):
-    print(prettify_string(inputs))
-
-
-def writelog(self, *logs, logdir="./logs", logfile="run.txt"):
-    f = os.path.join(logdir, logfile)
-    with open(f, "a+") as fh:
-        for log in logs:
-            fh.write(log)
-            fh.write("\n")
-
-
-def start_tensorboard(self, logdir="./logs"):
-    cmd = f"tensorboard --logdir={logdir} --port=6006"
-    os.system(cmd)
-
-
-def make_directory(save_dir):
-    if not os.path.exists(save_dir):
-        print(f"Making directory {save_dir}.")
-        os.mkdir(save_dir)
-
-
-def make_gif(filenames, duration, output_file=None):
-    try:
-        import imageio
-    except ModuleNotFoundError as e:
-        print("Please install Protoflow with [other] extra requirements.")
-        raise (e)
-
-    images = list()
-    for filename in filenames:
-        images.append(imageio.imread(filename))
-    if not output_file:
-        output_file = f"makegif.gif"
-    if images:
-        imageio.mimwrite(output_file, images, duration=duration)
-
-
-def gif_from_dir(directory,
-                 duration,
-                 prefix="",
-                 output_file=None,
-                 verbose=True):
-    images = os.listdir(directory)
-    if verbose:
-        print(f"Making gif from {len(images)} images under {directory}.")
-    filenames = list()
-    # Sort images
-    images = sorted(
-        images,
-        key=lambda img: int(os.path.splitext(img)[0].replace(prefix, "")))
-    for image in images:
-        fname = os.path.join(directory, image)
-        filenames.append(fname)
-    if not output_file:
-        output_file = os.path.join(directory, "makegif.gif")
-    make_gif(filenames=filenames, duration=duration, output_file=output_file)
-
-
-def accuracy_score(y_true, y_pred):
-    accuracy = np.sum(y_true == y_pred)
-    normalized_acc = accuracy / float(len(y_true))
-    return normalized_acc
-
-
-def predict_and_score(clf,
-                      x_test,
-                      y_test,
-                      verbose=False,
-                      title="Test accuracy"):
-    y_pred = clf.predict(x_test)
-    accuracy = np.sum(y_test == y_pred)
-    normalized_acc = accuracy / float(len(y_test))
-    if verbose:
-        print(f"{title}: {normalized_acc * 100:06.04f}%")
-    return normalized_acc
-
-
-def remove_nan_rows(arr):
-    """Remove all rows with `nan` values in `arr`."""
-    mask = np.isnan(arr).any(axis=1)
-    return arr[~mask]
-
-
-def remove_nan_cols(arr):
-    """Remove all columns with `nan` values in `arr`."""
-    mask = np.isnan(arr).any(axis=0)
-    return arr[~mask]
-
-
-def replace_in(arr, replacement_dict, inplace=False):
-    """Replace the keys found in `arr` with the values from
-    the `replacement_dict`.
-    """
-    if inplace:
-        new_arr = arr
-    else:
-        import copy
-
-        new_arr = copy.deepcopy(arr)
-    for k, v in replacement_dict.items():
-        new_arr[arr == k] = v
-    return new_arr
-
-
-def train_test_split(data, train=0.7, val=0.15, shuffle=None, return_xy=False):
-    """Split a classification dataset in such a way so as to
-    preserve the class distribution in subsamples of the dataset.
-    """
-    if train + val > 1.0:
-        raise ValueError("Invalid split values for train and val.")
-    Y = data[:, -1]
-    labels = set(Y)
-    hist = dict()
-    for l in labels:
-        data_l = data[Y == l]
-        nl = len(data_l)
-        nl_train = int(nl * train)
-        nl_val = int(nl * val)
-        nl_test = nl - (nl_train + nl_val)
-        hist[l] = (nl_train, nl_val, nl_test)
-
-    train_data = list()
-    val_data = list()
-    test_data = list()
-    for l, (nl_train, nl_val, nl_test) in hist.items():
-        data_l = data[Y == l]
-        if shuffle:
-            np.random.shuffle(data_l)
-        train_l = data_l[:nl_train]
-        val_l = data_l[nl_train:nl_train + nl_val]
-        test_l = data_l[nl_train + nl_val:nl_train + nl_val + nl_test]
-        train_data.append(train_l)
-        val_data.append(val_l)
-        test_data.append(test_l)
-
-    def _squash(data_list):
-        data = np.array(data_list[0])
-        for item in data_list[1:]:
-            data = np.vstack((data, np.array(item)))
-        return data
-
-    train_data = _squash(train_data)
-    if val_data:
-        val_data = _squash(val_data)
-    if test_data:
-        test_data = _squash(test_data)
-    if return_xy:
-        x_train = train_data[:, :-1]
-        y_train = train_data[:, -1]
-        x_val = val_data[:, :-1]
-        y_val = val_data[:, -1]
-        x_test = test_data[:, :-1]
-        y_test = test_data[:, -1]
-        return (x_train, y_train), (x_val, y_val), (x_test, y_test)
-    return train_data, val_data, test_data
-
-
-def class_histogram(data, title="Untitled"):
-    plt.figure(title)
-    plt.clf()
-    plt.title(title)
-    dist, counts = np.unique(data[:, -1], return_counts=True)
-    plt.bar(dist, counts)
-    plt.xticks(dist)
-    print("Call matplotlib.pyplot.show() to see the plot.")
-
-
-def ntimer(n=10):
-    """Wraps a function which wraps another function to time it."""
-    if n < 1:
-        raise (Exception(f"Invalid n = {n} given."))
-
-    def timer(func):
-        """Wraps `func` with a timer and returns the wrapped `func`."""
-        def wrapper(*args, **kwargs):
-            rv = None
-            before = time()
-            for _ in range(n):
-                rv = func(*args, **kwargs)
-            after = time()
-            elapsed = after - before
-            print(f"Elapsed: {elapsed*1e3:02.02f} ms")
-            return rv
-
-        return wrapper
-
-    return timer
-
-
-def memoize(verbose=True):
-    """Wraps a function which wraps another function that memoizes."""
-    def memoizer(func):
-        """Memoize (cache) return values of `func`.
-        Wraps `func` and returns the wrapped `func` so that `func`
-        is executed when the results are not available in the cache.
-        """
-        cache = {}
-
-        def wrapper(*args, **kwargs):
-            t = (pickle.dumps(args), pickle.dumps(kwargs))
-            if t not in cache:
-                if verbose:
-                    print(f"Adding NEW rv {func.__name__}{args}{kwargs} "
-                          "to cache.")
-                cache[t] = func(*args, **kwargs)
-            else:
-                if verbose:
-                    print(f"Using OLD rv {func.__name__}{args}{kwargs} "
-                          "from cache.")
-            return cache[t]
-
-        return wrapper
-
-    return memoizer

setup.py

@@ -20,6 +20,7 @@ INSTALL_REQUIRES = [
     "torch>=1.3.1",
     "torchvision>=0.5.0",
     "numpy>=1.9.1",
+    "sklearn",
 ]
 DATASETS = [
     "requests",
@@ -31,9 +32,9 @@ DOCS = [
     "sphinx",
     "sphinx_rtd_theme",
     "sphinxcontrib-katex",
+    "sphinx-autodoc-typehints",
 ]
 EXAMPLES = [
-    "sklearn",
     "matplotlib",
     "torchinfo",
 ]
@@ -42,7 +43,7 @@ ALL = DATASETS + DEV + DOCS + EXAMPLES + TESTS
 
 setup(
     name="prototorch",
-    version="0.4.1",
+    version="0.5.0",
     description="Highly extensible, GPU-supported "
     "Learning Vector Quantization (LVQ) toolbox "
     "built using PyTorch and its nn API.",

tests/test_components.py

@@ -0,0 +1,25 @@
+"""ProtoTorch components test suite."""
+
+import prototorch as pt
+import torch
+
+
+def test_labcomps_zeros_init():
+    protos = torch.zeros(3, 2)
+    c = pt.components.LabeledComponents(
+        distribution=[1, 1, 1],
+        initializer=pt.components.Zeros(2),
+    )
+    assert (c.components == protos).any() == True
+
+
+def test_labcomps_warmstart():
+    protos = torch.randn(3, 2)
+    plabels = torch.tensor([1, 2, 3])
+    c = pt.components.LabeledComponents(
+        distribution=[1, 1, 1],
+        initializer=None,
+        initialized_components=[protos, plabels],
+    )
+    assert (c.components == protos).any() == True
+    assert (c.component_labels == plabels).any() == True

tests/test_functions.py

@@ -4,9 +4,8 @@ import unittest
 
 import numpy as np
 import torch
-
 from prototorch.functions import (activations, competitions, distances,
-                                  initializers, losses)
+                                  initializers, losses, pooling)
 
 
 class TestActivations(unittest.TestCase):
@@ -105,10 +104,28 @@ class TestCompetitions(unittest.TestCase):
                                                         decimal=5)
         self.assertIsNone(mismatch)
 
+    def test_knnc_k1(self):
+        d = torch.tensor([[2.0, 3.0, 1.99, 3.01], [2.0, 3.0, 2.01, 3.0]])
+        labels = torch.tensor([0, 1, 2, 3])
+        actual = competitions.knnc(d, labels, k=1)
+        desired = torch.tensor([2, 0])
+        mismatch = np.testing.assert_array_almost_equal(actual,
+                                                        desired,
+                                                        decimal=5)
+        self.assertIsNone(mismatch)
+
+    def tearDown(self):
+        pass
+
+
+class TestPooling(unittest.TestCase):
+    def setUp(self):
+        pass
+
     def test_stratified_min(self):
         d = torch.tensor([[1.0, 0.0, 2.0, 3.0], [9.0, 8.0, 0, 1]])
         labels = torch.tensor([0, 0, 1, 2])
-        actual = competitions.stratified_min(d, labels)
+        actual = pooling.stratified_min_pooling(d, labels)
         desired = torch.tensor([[0.0, 2.0, 3.0], [8.0, 0.0, 1.0]])
         mismatch = np.testing.assert_array_almost_equal(actual,
                                                         desired,
@@ -119,28 +136,70 @@ class TestCompetitions(unittest.TestCase):
         d = torch.tensor([[1.0, 0.0, 2.0, 3.0], [9.0, 8.0, 0, 1]])
         labels = torch.tensor([0, 0, 1, 2])
         labels = torch.eye(3)[labels]
-        actual = competitions.stratified_min(d, labels)
+        actual = pooling.stratified_min_pooling(d, labels)
         desired = torch.tensor([[0.0, 2.0, 3.0], [8.0, 0.0, 1.0]])
         mismatch = np.testing.assert_array_almost_equal(actual,
                                                         desired,
                                                         decimal=5)
         self.assertIsNone(mismatch)
 
-    def test_stratified_min_simple(self):
+    def test_stratified_min_trivial(self):
         d = torch.tensor([[0.0, 2.0, 3.0], [8.0, 0, 1]])
         labels = torch.tensor([0, 1, 2])
-        actual = competitions.stratified_min(d, labels)
+        actual = pooling.stratified_min_pooling(d, labels)
         desired = torch.tensor([[0.0, 2.0, 3.0], [8.0, 0.0, 1.0]])
         mismatch = np.testing.assert_array_almost_equal(actual,
                                                         desired,
                                                         decimal=5)
         self.assertIsNone(mismatch)
 
-    def test_knnc_k1(self):
+    def test_stratified_max(self):
-        d = torch.tensor([[2.0, 3.0, 1.99, 3.01], [2.0, 3.0, 2.01, 3.0]])
+        d = torch.tensor([[1.0, 0.0, 2.0, 3.0, 9.0], [9.0, 8.0, 0, 1, 7.0]])
-        labels = torch.tensor([0, 1, 2, 3])
+        labels = torch.tensor([0, 0, 3, 2, 0])
-        actual = competitions.knnc(d, labels, k=torch.tensor([1]))
+        actual = pooling.stratified_max_pooling(d, labels)
-        desired = torch.tensor([2, 0])
+        desired = torch.tensor([[9.0, 3.0, 2.0], [9.0, 1.0, 0.0]])
+        mismatch = np.testing.assert_array_almost_equal(actual,
+                                                        desired,
+                                                        decimal=5)
+        self.assertIsNone(mismatch)
+
+    def test_stratified_max_one_hot(self):
+        d = torch.tensor([[1.0, 0.0, 2.0, 3.0, 9.0], [9.0, 8.0, 0, 1, 7.0]])
+        labels = torch.tensor([0, 0, 2, 1, 0])
+        labels = torch.nn.functional.one_hot(labels, num_classes=3)
+        actual = pooling.stratified_max_pooling(d, labels)
+        desired = torch.tensor([[9.0, 3.0, 2.0], [9.0, 1.0, 0.0]])
+        mismatch = np.testing.assert_array_almost_equal(actual,
+                                                        desired,
+                                                        decimal=5)
+        self.assertIsNone(mismatch)
+
+    def test_stratified_sum(self):
+        d = torch.tensor([[1.0, 0.0, 2.0, 3.0], [9.0, 8.0, 0, 1]])
+        labels = torch.LongTensor([0, 0, 1, 2])
+        actual = pooling.stratified_sum_pooling(d, labels)
+        desired = torch.tensor([[1.0, 2.0, 3.0], [17.0, 0.0, 1.0]])
+        mismatch = np.testing.assert_array_almost_equal(actual,
+                                                        desired,
+                                                        decimal=5)
+        self.assertIsNone(mismatch)
+
+    def test_stratified_sum_one_hot(self):
+        d = torch.tensor([[1.0, 0.0, 2.0, 3.0], [9.0, 8.0, 0, 1]])
+        labels = torch.tensor([0, 0, 1, 2])
+        labels = torch.eye(3)[labels]
+        actual = pooling.stratified_sum_pooling(d, labels)
+        desired = torch.tensor([[1.0, 2.0, 3.0], [17.0, 0.0, 1.0]])
+        mismatch = np.testing.assert_array_almost_equal(actual,
+                                                        desired,
+                                                        decimal=5)
+        self.assertIsNone(mismatch)
+
+    def test_stratified_prod(self):
+        d = torch.tensor([[1.0, 0.0, 2.0, 3.0, 9.0], [9.0, 8.0, 0, 1, 7.0]])
+        labels = torch.tensor([0, 0, 3, 2, 0])
+        actual = pooling.stratified_prod_pooling(d, labels)
+        desired = torch.tensor([[0.0, 3.0, 2.0], [504.0, 1.0, 0.0]])
         mismatch = np.testing.assert_array_almost_equal(actual,
                                                         desired,
                                                         decimal=5)

tests/test_modules.py

@@ -1,298 +0,0 @@
-"""ProtoTorch modules test suite."""
-
-import unittest
-
-import numpy as np
-import torch
-
-from prototorch.modules import losses, prototypes
-
-
-class TestPrototypes(unittest.TestCase):
-    def setUp(self):
-        self.x = torch.tensor(
-            [[0, -1, -2], [10, 11, 12], [0, 0, 0], [2, 2, 2]],
-            dtype=torch.float32)
-        self.y = torch.tensor([0, 0, 1, 1])
-        self.gen = torch.manual_seed(42)
-
-    def test_prototypes1d_init_without_input_dim(self):
-        with self.assertRaises(NameError):
-            _ = prototypes.Prototypes1D(nclasses=2)
-
-    def test_prototypes1d_init_without_nclasses(self):
-        with self.assertRaises(NameError):
-            _ = prototypes.Prototypes1D(input_dim=1)
-
-    def test_prototypes1d_init_with_nclasses_1(self):
-        with self.assertWarns(UserWarning):
-            _ = prototypes.Prototypes1D(nclasses=1, input_dim=1)
-
-    def test_prototypes1d_init_without_pdist(self):
-        p1 = prototypes.Prototypes1D(
-            input_dim=6,
-            nclasses=2,
-            prototypes_per_class=4,
-            prototype_initializer="ones",
-        )
-        protos = p1.prototypes
-        actual = protos.detach().numpy()
-        desired = torch.ones(8, 6)
-        mismatch = np.testing.assert_array_almost_equal(actual,
-                                                        desired,
-                                                        decimal=5)
-        self.assertIsNone(mismatch)
-
-    def test_prototypes1d_init_without_data(self):
-        pdist = [2, 2]
-        p1 = prototypes.Prototypes1D(input_dim=3,
-                                     prototype_distribution=pdist,
-                                     prototype_initializer="zeros")
-        protos = p1.prototypes
-        actual = protos.detach().numpy()
-        desired = torch.zeros(4, 3)
-        mismatch = np.testing.assert_array_almost_equal(actual,
-                                                        desired,
-                                                        decimal=5)
-        self.assertIsNone(mismatch)
-
-    def test_prototypes1d_proto_init_without_data(self):
-        with self.assertWarns(UserWarning):
-            _ = prototypes.Prototypes1D(
-                input_dim=3,
-                nclasses=2,
-                prototypes_per_class=1,
-                prototype_initializer="stratified_mean",
-                data=None,
-            )
-
-    def test_prototypes1d_init_torch_pdist(self):
-        pdist = torch.tensor([2, 2])
-        p1 = prototypes.Prototypes1D(input_dim=3,
-                                     prototype_distribution=pdist,
-                                     prototype_initializer="zeros")
-        protos = p1.prototypes
-        actual = protos.detach().numpy()
-        desired = torch.zeros(4, 3)
-        mismatch = np.testing.assert_array_almost_equal(actual,
-                                                        desired,
-                                                        decimal=5)
-        self.assertIsNone(mismatch)
-
-    def test_prototypes1d_init_without_inputdim_with_data(self):
-        _ = prototypes.Prototypes1D(
-            nclasses=2,
-            prototypes_per_class=1,
-            prototype_initializer="stratified_mean",
-            data=[[[1.0], [0.0]], [1, 0]],
-        )
-
-    def test_prototypes1d_init_with_int_data(self):
-        _ = prototypes.Prototypes1D(
-            nclasses=2,
-            prototypes_per_class=1,
-            prototype_initializer="stratified_mean",
-            data=[[[1], [0]], [1, 0]],
-        )
-
-    def test_prototypes1d_init_one_hot_without_data(self):
-        _ = prototypes.Prototypes1D(
-            input_dim=1,
-            nclasses=2,
-            prototypes_per_class=1,
-            prototype_initializer="stratified_mean",
-            data=None,
-            one_hot_labels=True,
-        )
-
-    def test_prototypes1d_init_one_hot_labels_false(self):
-        """Test if ValueError is raised when `one_hot_labels` is set to `False`
-        but the provided `data` has one-hot encoded labels.
-        """
-        with self.assertRaises(ValueError):
-            _ = prototypes.Prototypes1D(
-                input_dim=1,
-                nclasses=2,
-                prototypes_per_class=1,
-                prototype_initializer="stratified_mean",
-                data=([[0.0], [1.0]], [[0, 1], [1, 0]]),
-                one_hot_labels=False,
-            )
-
-    def test_prototypes1d_init_1d_y_data_one_hot_labels_true(self):
-        """Test if ValueError is raised when `one_hot_labels` is set to `True`
-        but the provided `data` does not contain one-hot encoded labels.
-        """
-        with self.assertRaises(ValueError):
-            _ = prototypes.Prototypes1D(
-                input_dim=1,
-                nclasses=2,
-                prototypes_per_class=1,
-                prototype_initializer="stratified_mean",
-                data=([[0.0], [1.0]], [0, 1]),
-                one_hot_labels=True,
-            )
-
-    def test_prototypes1d_init_one_hot_labels_true(self):
-        """Test if ValueError is raised when `one_hot_labels` is set to `True`
-        but the provided `data` contains 2D targets but
-        does not contain one-hot encoded labels.
-        """
-        with self.assertRaises(ValueError):
-            _ = prototypes.Prototypes1D(
-                input_dim=1,
-                nclasses=2,
-                prototypes_per_class=1,
-                prototype_initializer="stratified_mean",
-                data=([[0.0], [1.0]], [[0], [1]]),
-                one_hot_labels=True,
-            )
-
-    def test_prototypes1d_init_with_int_dtype(self):
-        with self.assertRaises(RuntimeError):
-            _ = prototypes.Prototypes1D(
-                nclasses=2,
-                prototypes_per_class=1,
-                prototype_initializer="stratified_mean",
-                data=[[[1], [0]], [1, 0]],
-                dtype=torch.int32,
-            )
-
-    def test_prototypes1d_inputndim_with_data(self):
-        with self.assertRaises(ValueError):
-            _ = prototypes.Prototypes1D(input_dim=1,
-                                        nclasses=1,
-                                        prototypes_per_class=1,
-                                        data=[[1.0], [1]])
-
-    def test_prototypes1d_inputdim_with_data(self):
-        with self.assertRaises(ValueError):
-            _ = prototypes.Prototypes1D(
-                input_dim=2,
-                nclasses=2,
-                prototypes_per_class=1,
-                prototype_initializer="stratified_mean",
-                data=[[[1.0], [0.0]], [1, 0]],
-            )
-
-    def test_prototypes1d_nclasses_with_data(self):
-        """Test ValueError raise if provided `nclasses` is not the same
-        as the one computed from the provided `data`.
-        """
-        with self.assertRaises(ValueError):
-            _ = prototypes.Prototypes1D(
-                input_dim=1,
-                nclasses=1,
-                prototypes_per_class=1,
-                prototype_initializer="stratified_mean",
-                data=[[[1.0], [2.0]], [1, 2]],
-            )
-
-    def test_prototypes1d_init_with_ppc(self):
-        p1 = prototypes.Prototypes1D(data=[self.x, self.y],
-                                     prototypes_per_class=2,
-                                     prototype_initializer="zeros")
-        protos = p1.prototypes
-        actual = protos.detach().numpy()
-        desired = torch.zeros(4, 3)
-        mismatch = np.testing.assert_array_almost_equal(actual,
-                                                        desired,
-                                                        decimal=5)
-        self.assertIsNone(mismatch)
-
-    def test_prototypes1d_init_with_pdist(self):
-        p1 = prototypes.Prototypes1D(
-            data=[self.x, self.y],
-            prototype_distribution=[6, 9],
-            prototype_initializer="zeros",
-        )
-        protos = p1.prototypes
-        actual = protos.detach().numpy()
-        desired = torch.zeros(15, 3)
-        mismatch = np.testing.assert_array_almost_equal(actual,
-                                                        desired,
-                                                        decimal=5)
-        self.assertIsNone(mismatch)
-
-    def test_prototypes1d_func_initializer(self):
-        def my_initializer(*args, **kwargs):
-            return torch.full((2, 99), 99.0), torch.tensor([0, 1])
-
-        p1 = prototypes.Prototypes1D(
-            input_dim=99,
-            nclasses=2,
-            prototypes_per_class=1,
-            prototype_initializer=my_initializer,
-        )
-        protos = p1.prototypes
-        actual = protos.detach().numpy()
-        desired = 99 * torch.ones(2, 99)
-        mismatch = np.testing.assert_array_almost_equal(actual,
-                                                        desired,
-                                                        decimal=5)
-        self.assertIsNone(mismatch)
-
-    def test_prototypes1d_forward(self):
-        p1 = prototypes.Prototypes1D(data=[self.x, self.y])
-        protos, _ = p1()
-        actual = protos.detach().numpy()
-        desired = torch.ones(2, 3)
-        mismatch = np.testing.assert_array_almost_equal(actual,
-                                                        desired,
-                                                        decimal=5)
-        self.assertIsNone(mismatch)
-
-    def test_prototypes1d_dist_validate(self):
-        p1 = prototypes.Prototypes1D(input_dim=0, prototype_distribution=[0])
-        with self.assertWarns(UserWarning):
-            _ = p1._validate_prototype_distribution()
-
-    def test_prototypes1d_validate_extra_repr_not_empty(self):
-        p1 = prototypes.Prototypes1D(input_dim=0, prototype_distribution=[0])
-        rep = p1.extra_repr()
-        self.assertNotEqual(rep, "")
-
-    def tearDown(self):
-        del self.x, self.y, self.gen
-        _ = torch.seed()
-
-
-class TestLosses(unittest.TestCase):
-    def setUp(self):
-        pass
-
-    def test_glvqloss_init(self):
-        _ = losses.GLVQLoss(0, "swish_beta", beta=20)
-
-    def test_glvqloss_forward_1ppc(self):
-        criterion = losses.GLVQLoss(margin=0,
-                                    squashing="sigmoid_beta",
-                                    beta=100)
-        d = torch.stack([torch.ones(100), torch.zeros(100)], dim=1)
-        labels = torch.tensor([0, 1])
-        targets = torch.ones(100)
-        outputs = [d, labels]
-        loss = criterion(outputs, targets)
-        loss_value = loss.item()
-        self.assertAlmostEqual(loss_value, 0.0)
-
-    def test_glvqloss_forward_2ppc(self):
-        criterion = losses.GLVQLoss(margin=0,
-                                    squashing="sigmoid_beta",
-                                    beta=100)
-        d = torch.stack([
-            torch.ones(100),
-            torch.ones(100),
-            torch.zeros(100),
-            torch.ones(100)
-        ],
-                        dim=1)
-        labels = torch.tensor([0, 0, 1, 1])
-        targets = torch.ones(100)
-        outputs = [d, labels]
-        loss = criterion(outputs, targets)
-        loss_value = loss.item()
-        self.assertAlmostEqual(loss_value, 0.0)
-
-    def tearDown(self):
-        pass