If you look at this GitHub link ( here is the paper link for the implementation ) you can see that the get_train_instances method generates trainingns instances. In addition, negative test data is loaded.

My question now is why do I create such training instances and why do I have to load negative data?

Are there any other scientific publications that describe this in more detail? I would be very happy if you could help me with a source.

  • $\begingroup$ I'm not familiar with this work in specific, but in general, negative sampling is a concept that appears in word embeddings, because the target vector is huge in scale (i.e. its size is the number of unique words in the dataset). It allows you to update a small percentage of weights instead of all of them with each training sample. I don't know if this helps in your case... $\endgroup$ – Djib2011 Jan 4 at 10:55

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.