Getting started#
PyKale Tutorials are runnable scripts. You read the narrated version, then run the full experiment when you want the real numbers. Here is the shortest path in.
Install PyKale#
pip install pykale
Datasets are never bundled — each tutorial downloads what it needs, or points at pykale/data.
Run your first tutorial#
The recommended starting point is domain adaptation on digit images. It trains a classifier on MNIST and adapts it to USPS without target labels — PyKale’s core idea in a few minutes on CPU.
cd others/image_classification/digit_domain_adaptation
pip install -r requirements.txt
python main.py --cfg configs/MN2UP-DANN.yaml --devices 1
Explore a domain#
Tutorials are filed by the domain that owns them:
Drug-target binding and polypharmacy side effects.
Multiomics integration for cancer classification.
Cardiac imaging and uncertainty estimation.
Method demonstrations: recognition, domain adaptation, few-shot learning.
Reserved — tutorials in development.
Adapt an example to your own data#
Every tutorial ships its complete main.py and config. Fork the folder, point
DATASET.ROOT (or the equivalent config field) at your data, and adjust the model. See
Contributing for conventions, and All tutorials to browse everything.