Arrange the following steps in a proper sequence for the process of training a…

Correct sequence: Weight initialization → Feed forward → Loss calculation → Back propagation → Weight update.

1. Weight initialization: set initial values for model parameters (weights and biases).

2. Feed forward: pass inputs through the network to compute predictions (activations).

3. Loss calculation: compute the loss/error between predictions and true targets.

4. Back propagation: compute gradients of the loss with respect to model parameters using the forward-pass activations.

5. Weight update: update weights using the computed gradients (for example, a gradient descent step).

Repeat these steps for each mini-batch or epoch until the model converges.

Why this order matters:

• You cannot compute loss without predictions from the forward pass.

• Backpropagation requires the loss and intermediate activations to compute gradients correctly.

• Weight updates must use gradients from backpropagation; updating before computing gradients is invalid.