Best Practice:
The best approach is to cite both Keras *and* the backend you used. This provides the most complete picture of your computational environment.
* For TensorFlow backend: Cite TensorFlow and then Keras. The TensorFlow citation will vary depending on the version you used. Check the TensorFlow website for the most up-to-date citation information. Then, you'd cite Keras as below.
* For other backends (if applicable): Replace TensorFlow with the appropriate backend (e.g., Theano, CNTK) and its corresponding citation.
Keras Citation (to be used in conjunction with the backend citation):
You should cite the Keras paper:
> Chollet, F. et al. (2015). Keras. https://keras.io
Example Citations (choose the one that matches your setup):
Example 1 (TensorFlow backend):
* [TensorFlow citation] — This needs to be the specific TensorFlow citation for the version you used (find this on the TensorFlow website)
* Chollet, F. et al. (2015). Keras. https://keras.io
Example 2 (using a specific version of TensorFlow):
* Abdi, A., et al. (Year). TensorFlow: Large-Scale Machine Learning on Heterogeneous Distributed Systems. *(Provide full citation details as found on the TensorFlow website for the relevant version)*
* Chollet, F. et al. (2015). Keras. https://keras.io
Important Considerations:
* Version Numbers: For reproducibility, consider including the specific versions of Keras and the backend you used (e.g., "Keras v2.11.0 with TensorFlow v2.12.0"). You can include this information in the text or a supplementary materials section.
* BibTeX: Most bibliographic management software (like Zotero or Mendeley) allows you to create BibTeX entries. You might need to manually create one for the Keras link if your software doesn't automatically find it.
Remember to always consult the style guide of your target publication for specific formatting requirements.