In word embedding models like Word2Vec, what does the 'Skip-gram' architecture aim to do?

1. ক
   Predict the current word based on its context words
   
2. খ
   Predict context words based on the current word
   
3. গ
   Map words to a low-dimensional space using matrix factorization
   
4. ঘ
   Calculate the frequency of words in a corpus