Hi, I’m wondering what the meanings of terms b_i and b’_j that appear in the video around 6:30 are? Andrew added these terms on the slide but I couldn’t find where he explained where these two terms came from. Thanks!
Hello @Kuang-Han_Huang,
Each word carries a trainable embedding vector and a trainable bias and b means that bias. If you want to dig deeper, you might read the GloVe paper for its section 3.
Cheers,
Raymond
1 Like
Thanks so much @rmwkwok !
You are welcome, @Kuang-Han_Huang