目次
Features of fasttext
- Improved objective function
- Consideration of negative samples
- This should not affect training time.
- Change in optimization method
- Use of stochastic optimization
- If it affects the learning time, it should be this one
- Implementation in C language
- This is the most effective, isn’t it?
If we implement it in ## pytorch, it won’t be much different from word2vec. It would depend on the amount of data to be trained.
It depends on the amount of data to be trained, but it is unlikely that a toy program can measure a significant difference.
Are there hidden implementation innovations that are not described in the paper?
- Is there an implementation device?
- Isn’t it unfair not to describe it in the paper?
- I think it is dishonest to say that the implementation made it faster.
I can’t tell what is fast just by reading the paper.
- I wonder what makes it fast?
Translated with www.DeepL.com/Translator (free version)
See also
- Creating data in Natural Language Inference (NLI) format for Sentence transformer
- On the use of distributed representations bagging for class classification and generalization performance
- How to train a Japanese model with Sentence transformer to get a distributed representation of a sentence
- Using BART (sentence summary model) with hugging face
- Procedure for obtaining a distributed representation of a Japanese sentence using a trained Universal Sentence Encoder