GRU forward pass - ML & AI Coding

217. GRU forward pass

medium

General

staff

Implement the forward pass of a single-layer GRU to process a sequence of input vectors and produce the sequence of hidden states. You’ll compute the GRU gates and hidden updates step-by-step using the standard GRU equations.

A GRU update at time step (t) is defined as:

\begin{aligned} z_t &= \sigma(W_z x_t + U_z h_{t-1} + b_z) \\ r_t &= \sigma(W_r x_t + U_r h_{t-1} + b_r) \\ \tilde{h}_t &= \tanh(W_h x_t + U_h (r_t \odot h_{t-1}) + b_h) \\ h_t &= (1 - z_t)\odot h_{t-1} + z_t \odot \tilde{h}_t \end{aligned}

Requirements

Implement the function

python

Rules:

Treat each x_t as a row from X and compute h_t sequentially from t=0 to T-1.
Implement sigmoid and tanh yourself using NumPy (no deep learning frameworks).
Use elementwise multiply for (\odot) and keep the math consistent with the formulas above.
Return the full set of hidden states (not just the last one) as a NumPy array.
Keep everything inside this single function (helper inner functions are fine).

Example

python

Output:

python

Input Signature

Argument	Type
X	np.ndarray
h0	np.ndarray
params	dict

Output Signature

Return Name	Type
value	np.ndarray

Constraints

Use only NumPy; no deep learning frameworks.
Matrix shapes must match specified dimensions.
Return np.ndarray of T hidden states.

Hint 1

Convert X, h0, and all params entries to NumPy arrays first so you can use @ for matrix–vector products and * for elementwise multiply.

Hint 2

Inside a loop over timesteps, compute the gates in the exact order: z_t = sigmoid(Wz@x_t + Uz@h + bz) and r_t = sigmoid(Wr@x_t + Ur@h + br); watch shapes: x_t is (input_dim,), h is (hidden_dim,).

Hint 3

Compute h_tilde = tanh(Wh@x_t + Uh@(r_t*h) + bh), then update: h = (1-z_t)*h + z_t*h_tilde. Append a copy (e.g., .tolist()) each step to return all hidden states.

Roles

ML Engineer

AI Engineer

Companies

General

Levels

staff

senior

entry

Input Arguments

Edit values below to test with custom inputs

You need tolog in/sign upto run or submit