FAQ

1. I just published a paper with a new algorithm. Can you add it to the library?

Absolutely! Please reach out directly or open a new issue in this repository.

2. I like algorithm XY a lot, but cannot find it implemented in salad

Please also reach out or open a new issue in this repository. We had to make a choice of which algorithms to include and are happy to extend the library, even if an algorithm does not meet the state of the art anymore. One central goal is the applicability to new datasets, and here it is unclear which algorithm will actually become state of the art.

3. I want to adapt an already trained model to a new dataset and don't see how this is possible

We are working on adding functionality for these use cases. In general, one easy thing to try is the adaptation of batch-norm statistics, followed by supervised finetuning. Have a look at salad.structural for some example scripts for these use cases. We are planning to add code for doing this in unsupervised ways in the future. If you really need this functionality, feel free to open an issue or comment on existing ones!

4. I cannot find my question in this FAQ.

Please open an issue or reach out! We will extend the FAQ from time to time.