The following statement about recurrent neural networks is wrong?

Question

Options

• ARecurrent neural network can be unfolded according to the time axis
• BLSTM Unable to solve the problem of vanishing gradient
• CLSTM It is also a recurrent neural network
• DRecurrent neural network can be abbreviated as RNN

• A
  9% (2)
• B
  83% (19)
• C
  4% (1)
• D
  4% (1)

Explanation

Option B is the false statement because LSTM (Long Short-Term Memory) was specifically designed to solve the vanishing gradient problem. Its gating mechanisms - input, forget, and output gates - along with a cell state allow gradients to flow through many time steps without vanishing, which is the core architectural innovation that makes LSTMs more powerful than vanilla RNNs for long sequences.

The other three statements are all true, which is why they are distractors: RNNs can indeed be unfolded along the time axis (A) to visualize how the network processes sequential data step by step; LSTM is absolutely a type of recurrent neural network (C), just an enhanced one; and RNN is the standard abbreviation for Recurrent Neural Network (D).

Memory tip: Remember that LSTM was invented because vanilla RNNs suffered from vanishing gradients - so by definition, LSTM must address that problem. Any statement saying LSTM cannot solve it is the opposite of its entire purpose.

No community discussion yet for this question.