Learning Phrase Representations: The GRU Update Gate
Beginner
The problem, from first principles
A gated recurrent unit decides how much of its old state to keep and how much to replace. The update gate is the first decision: it is a vector of values between 0 and 1 computed from the current input and the previous state.
From theory to code
Implement gru_update_gate(x, h, Wz, Uz), which returns sigmoid(Wz x + Uz h).
Constraints
- Output values are strictly between 0 and 1.
Hints
Hint 1
Compute the affine map of x and h, then apply the logistic function element-wise.
Keywords
research-paperstransformersllmrecurrentgru
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.