Memory Networks: Reading the Output Memory
Beginner
The problem, from first principles
After attention chooses which memories matter, the model reads an output representation of each memory, weighted by that attention. The sum is the evidence the answer is built from.
From theory to code
Implement memory_output(p, outputs), the attention-weighted sum of output embeddings.
Constraints
- Weights sum to one.
Hints
Hint 1
Multiply the attention vector by the output matrix.
Keywords
research-paperssequence-modelsmemoryquestion-answering
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.