# `datasets`

Use built-in datasets for online continual learning.

In OCL datastreams are irreversible sequences of examples following a
non-stationary data distribution. Learners in OCL can only learn from a single
pass through the datastream but are expected to perform well on any portion of
the datastream.

Portions of the datastream where the data distribution is relatively stationary
are called *tasks*.

A common way to construct an OCL dataset for experimentation is to group the
classes of a classification dataset into tasks. Known as the *class-incremental*
scenario, the learner is presented with a sequence of tasks where each task
contains a new subset of the classes.

For example [`SplitMNIST`](capymoa.ocl.datasets.SplitMNIST.md#capymoa.ocl.datasets.SplitMNIST) splits the MNIST dataset into five tasks where
each task contains two classes:

```pycon
>>> from capymoa.ocl.datasets import SplitMNIST
>>> scenario = SplitMNIST(preload_test=False)
>>> scenario.task_schedule
[{1, 4}, {5, 7}, {9, 3}, {0, 8}, {2, 6}]
```

To get the usual CapyMOA stream object for training:

```pycon
>>> instance = scenario.stream.next_instance()
>>> instance
LabeledInstance(
    Schema(SplitMNIST10/5),
    x=[0. 0. 0. ... 0. 0. 0.],
    y_index=4,
    y_label='4'
)
```

CapyMOA streams flatten the data into a feature vector:

```pycon
>>> instance.x.shape
(784,)
```

You can access the PyTorch datasets for each task:

```pycon
>>> x, y = scenario.test_tasks[0][0]
>>> x.shape
torch.Size([1, 28, 28])
>>> y
1
```

## Classes

| [`DomainCIFAR100`](capymoa.ocl.datasets.DomainCIFAR100.md#capymoa.ocl.datasets.DomainCIFAR100)           | Domain incremental CIFAR-100 variant with 20 classes per task.                  |
|---------------------------------------------------------------------------------------------------------------|---------------------------------------------------------------------------------|
| [`DomainCIFAR100ViT`](capymoa.ocl.datasets.DomainCIFAR100ViT.md#capymoa.ocl.datasets.DomainCIFAR100ViT)     | Domain incremental CIFAR-100 ViT variant with 20 classes per task.              |
| [`RotatedFashionMNIST`](capymoa.ocl.datasets.RotatedFashionMNIST.md#capymoa.ocl.datasets.RotatedFashionMNIST) | Domain-incremental FashionMNIST where each task applies a fixed image rotation. |
| [`RotatedMNIST`](capymoa.ocl.datasets.RotatedMNIST.md#capymoa.ocl.datasets.RotatedMNIST)               | Rotated MNIST where each task applies a fixed image rotation.                   |
| [`RotatedTinyMNIST`](capymoa.ocl.datasets.RotatedTinyMNIST.md#capymoa.ocl.datasets.RotatedTinyMNIST)       | Domain-incremental TinyMNIST where each task applies a fixed image rotation.    |
| [`SplitCIFAR10`](capymoa.ocl.datasets.SplitCIFAR10.md#capymoa.ocl.datasets.SplitCIFAR10)               | Split CIFAR-10 dataset for online class incremental learning.                   |
| [`SplitCIFAR10ViT`](capymoa.ocl.datasets.SplitCIFAR10ViT.md#capymoa.ocl.datasets.SplitCIFAR10ViT)         | CIFAR10 encoded by a Vision Transformer (ViT).                                  |
| [`SplitCIFAR100`](capymoa.ocl.datasets.SplitCIFAR100.md#capymoa.ocl.datasets.SplitCIFAR100)             | Split CIFAR-100 dataset for online class incremental learning.                  |
| [`SplitCIFAR100ViT`](capymoa.ocl.datasets.SplitCIFAR100ViT.md#capymoa.ocl.datasets.SplitCIFAR100ViT)       | CIFAR100 encoded by a Vision Transformer (ViT).                                 |
| [`SplitFashionMNIST`](capymoa.ocl.datasets.SplitFashionMNIST.md#capymoa.ocl.datasets.SplitFashionMNIST)     | Split Fashion MNIST dataset for online class incremental learning.              |
| [`SplitMNIST`](capymoa.ocl.datasets.SplitMNIST.md#capymoa.ocl.datasets.SplitMNIST)                   | Split MNIST dataset for online class incremental learning.                      |
| [`TinySplitMNIST`](capymoa.ocl.datasets.TinySplitMNIST.md#capymoa.ocl.datasets.TinySplitMNIST)           | A lower resolution and smaller version of the SplitMNIST dataset for testing.   |