Neural Network Recurrent Neural Networks

Question 1

What makes recurrent neural networks (RNNs) different from feedforward networks?

Accepted Answer

They maintain a hidden state that carries information across time steps

Answer

They use convolutional operations

Answer

They do not use backpropagation

Answer

They process inputs in random order

Question 2

What does LSTM stand for and what problem was it designed to solve?

Accepted Answer

Long Short-Term Memory; vanishing/exploding gradients in standard RNNs

Answer

Large Scale Temporal Model; slow training in deep networks

Answer

Layered Sequential Training Module; overfitting in time-series models

Answer

Learned State Transition Memory; sparse gradient updates

Question 3

What are the three gates in an LSTM cell?

Accepted Answer

Forget, input, and output gates

Answer

Input, process, output gates

Answer

Read, write, and erase gates

Answer

Encode, decode, and attention gates

Question 4

What is a bidirectional RNN?

Accepted Answer

An RNN that processes data both forwards and backwards in time simultaneously

Answer

An RNN that can handle both text and image inputs

Answer

An RNN that uses two separate hidden states for long and short term

Answer

An RNN where gradients flow in two directions during backpropagation

Question 5

What is Backpropagation Through Time (BPTT)?

Accepted Answer

The algorithm for computing gradients in RNNs by unrolling the network across time steps

Answer

A technique to speed up RNN training using time-based learning rates

Answer

A method for transferring RNN weights across time periods

Answer

A regularization technique specific to sequential models

Question 6

What is the key difference between GRU and LSTM?

Accepted Answer

GRU has two gates instead of three and no separate cell state, making it simpler

Answer

GRU has no gating mechanism

Answer

GRU uses multiplicative interactions instead of additive

Answer

GRU cannot handle long-term dependencies