Adversarial domain adaptation methods.

This example illustrates the adversarial methods from on a simple image classification task.

# Author: Théo Gnassounou
#
# License: BSD 3-Clause
# sphinx_gallery_thumbnail_number = 4

from skorch import NeuralNetClassifier
from torch import nn

from skada.datasets import load_mnist_usps
from skada.deep import DANN
from skada.deep.modules import MNISTtoUSPSNet

Load the image datasets

dataset = load_mnist_usps(n_classes=2, n_samples=0.5, return_dataset=True)
X, y, sample_domain = dataset.pack(
    as_sources=["mnist"], as_targets=["usps"], mask_target_labels=True
)
X_test, y_test, sample_domain_test = dataset.pack(
    as_sources=[], as_targets=["usps"], mask_target_labels=False
)

/home/circleci/project/skada/datasets/_mnist_usps.py:72: UserWarning: To copy construct from a tensor, it is recommended to use sourceTensor.detach().clone() or sourceTensor.detach().clone().requires_grad_(True), rather than torch.tensor(sourceTensor).
  mnist_target = torch.tensor(mnist_dataset.targets)

Train a classic model

model = NeuralNetClassifier(
    MNISTtoUSPSNet(),
    criterion=nn.CrossEntropyLoss(),
    batch_size=128,
    max_epochs=5,
    train_split=False,
    lr=1e-2,
)
model.fit(X[sample_domain > 0], y[sample_domain > 0])
model.score(X_test, y=y_test)

  epoch    train_loss      dur
-------  ------------  -------
      1        1.7794  12.5957
      2        0.4562  10.5982
      3        0.1501  11.7958
      4        0.0869  7.8996
      5        0.0568  5.5026

0.9260450160771704

Train a DANN model

model = DANN(
    MNISTtoUSPSNet(),
    layer_name="fc1",
    batch_size=128,
    max_epochs=5,
    train_split=False,
    reg=0.01,
    num_features=128,
    lr=1e-2,
)
model.fit(X, y, sample_domain=sample_domain)
model.score(X_test, y_test, sample_domain=sample_domain_test)

/home/circleci/.local/lib/python3.10/site-packages/torch/nn/modules/module.py:1773: UserWarning: Implicit dimension choice for softmax has been deprecated. Change the call to include dim=X as an argument.
  return self._call_impl(*args, **kwargs)
  epoch    train_loss      dur
-------  ------------  -------
      1        2.6083  16.5485
      2        1.3799  14.8978
      3        1.1277  14.0039
      4        1.0700  27.8362
      5        1.0587  18.1998

0.9742765273311897