Which architecture is especially suitable for modeling sequential data and maintaining context across steps?

Which architecture is especially suitable for modeling sequential data and maintaining context across steps?

• A. Transformer
• B. Convolutional Neural Network
• C. Recurrent Neural Network
• D. Generative Adversarial Network

Answer: (C)

Modeling sequences requires memory of what happened earlier. A recurrent neural network provides that by carrying a hidden state through the sequence, so each step combines the current input with the previous context to produce an output and an updated context. This setup lets the model handle variable-length sequences and learn dependencies that span many steps, which is exactly what you need when what comes next depends on past elements. Training through time adjusts how much influence earlier steps have on later predictions, reinforcing long-range consistency. Other architectures have strengths—convolutional networks focus on local patterns, transformers use attention to relate many positions in a sequence, and GANs aren’t built for sequential dependencies—but they don’t naturally maintain step-to-step context in the same flowing, memory-based way.

Modeling sequences requires memory of what happened earlier. A recurrent neural network provides that by carrying a hidden state through the sequence, so each step combines the current input with the previous context to produce an output and an updated context. This setup lets the model handle variable-length sequences and learn dependencies that span many steps, which is exactly what you need when what comes next depends on past elements. Training through time adjusts how much influence earlier steps have on later predictions, reinforcing long-range consistency. Other architectures have strengths—convolutional networks focus on local patterns, transformers use attention to relate many positions in a sequence, and GANs aren’t built for sequential dependencies—but they don’t naturally maintain step-to-step context in the same flowing, memory-based way.