I'd assume that a) is false since for GloVe, we need to first obtain the word cooccurrence counts which require us to first make a pass over the whole input text. Only after this is done, we can apply SGD for GloVe but not earlier, i.e. it doesn't really fit in the streaming model.
I'd assume that a) is false since for GloVe, we need to first obtain the word cooccurrence counts which require us to first make a pass over the whole input text. Only after this is done, we can apply SGD for GloVe but not earlier, i.e. it doesn't really fit in the streaming model.
No, word2vec doesn't need the co-occurrence matrix (and thus would be applicable in the streaming model). You can find the details in the Section 3 of the original paper.
Pb 21 exam 2017
Hello,
Why answer a) is false for the question 21 please?
2
I'd assume that a) is false since for GloVe, we need to first obtain the word cooccurrence counts which require us to first make a pass over the whole input text. Only after this is done, we can apply SGD for GloVe but not earlier, i.e. it doesn't really fit in the streaming model.
1
I don't know the answer to your question, but let me post the question here so others can answer it more easily.
3
Interested in this question too and why c is false also please?
1
For C, as can be referred from the slide that FastText is supervised.
I'd assume that a) is false since for GloVe, we need to first obtain the word cooccurrence counts which require us to first make a pass over the whole input text. Only after this is done, we can apply SGD for GloVe but not earlier, i.e. it doesn't really fit in the streaming model.
1
Does word2vec also need the co occurence matrix ?
No, word2vec doesn't need the co-occurrence matrix (and thus would be applicable in the streaming model). You can find the details in the Section 3 of the original paper.
Add comment