Luong Attention: The General Score
Intermediate
The problem, from first principles
The general score adds a learned matrix between the two states. It lets the decoder and encoder live in different spaces and still be compared.
From theory to code
Implement general_score(h_t, W, h_s), the bilinear alignment score.
Constraints
- Returns a float.
Hints
Hint 1
Multiply the decoder state by the matrix, then dot with the encoder state.
Keywords
research-paperssequence-modelsattentiontranslation
Python
Run your code to see results
No results yet
Run your code to check it against the examples, or submit to run every test.