Which of the following is not included in the recurrent neural network usage scenario?

Question

Options

• AMachine translation
• BSpeech recognition
• CImage style migration
• DText generation

• A
  5% (2)
• B
  7% (3)
• C
  71% (30)
• D
  17% (7)

Explanation

Image style migration (neural style transfer) is a task that operates on spatial pixel data - it requires capturing local patterns, textures, and features across a 2D image. That's the domain of Convolutional Neural Networks (CNNs), which use spatial filters, not sequence modeling.

RNNs are designed for sequential, time-dependent data - they maintain a hidden state across time steps to capture context. Machine translation (A) feeds a sequence of words into an encoder RNN and decodes it into another language sequence. Speech recognition (B) processes audio frames as a time series, exactly the kind of sequential input RNNs excel at. Text generation (D) predicts the next token based on previous context - a classic RNN application (and the foundation of early language models).

Memory tip: Think "R for Recurrent = R for Running through time." If the task involves something that flows over time or in order (words, audio, sentences), it's RNN territory. Images don't have a natural sequential order, so image tasks belong to CNNs - not RNNs.

No community discussion yet for this question.