Which type of neural networks is designed to handle sequential data more effectively than traditional neural networks?

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

Answer: (B)

Sequential data requires remembering what came before to make sense of what comes next. Recurrent neural networks incorporate feedback connections that pass the hidden state from one step to the next, creating a memory of prior inputs. This ability to carry information through time lets the network model temporal dependencies and context, which is essential for tasks like language, time series, and any sequence where order matters. Training them with backpropagation through time adjusts these recurrent connections so the network learns how past information influences current predictions.

Convolutional networks excel at detecting patterns in spatial or local temporal structure but don’t maintain a state across arbitrary-length sequences by default. Transformers use self-attention to relate all parts of the sequence, effectively modeling dependencies without recurrence, but that’s a different approach to sequence handling that doesn’t rely on the same internal memory loop. Generative adversarial networks focus on creating data distributions rather than modeling sequential dependencies over time.

Since the question targets a design that inherently handles sequential data with memory across time, the recurrent neural network is the best fit.