Jeffrey Pennington

Glove: Global Vectors for Word Representation
Jeffrey Pennington, Richard Socher and Christopher Manning

[ting-hsiangwang]

Thank you. Now I finally understand the reason behind the statement "Since vector spaces are inherently linear structures, the most natural way to do this (encoding) is with vector differences." in the original paper. 6:00

[khalednaami]

Hi,
Thanks for the nice presentation. At 9:51, shouldn't argmax formula has a ||Wx|| in the denominator?
Thank you,