Recurrent Neural Networks — Checkpoint Quiz

A is correct. The defining feature of an RNN is its hidden state, which carries information from earlier time steps forward, letting the network's output at time t depend on everything it has seen so far, not just the current input alone.

A is correct. Backpropagation through time unrolls the recurrence and multiplies gradients across every time step. Repeated multiplication by the same matrix tends to shrink (vanish) or grow (explode) the gradient exponentially with sequence length, making long-range dependencies hard to learn.

A is correct. Unrolling represents the RNN's repeated application across time steps as one long computational graph, with the same weights reused at each step. This is what makes it possible to apply ordinary backpropagation (called backpropagation through time) to compute gradients.