DecodeAI
← Question Bank

Deep Learning & Generative AI

Sequence Modelling Interview Questions and Answers

Interview questions on Sequence Modelling Interview Questions and Answers.

9 questions

Recurrent Neural Networks (RNNs)

Q1. What is the primary motivation for using Recurrent Neural Networks (RNNs) over standard Feedforward Neural Networks for sequential data?

Sign in to bookmark

Recurrent Neural Networks (RNNs)

Q2. What are the main limitations of basic (vanilla) RNNs?

Sign in to bookmark

Gated Architectures (LSTM & GRU)

Q3. Explain the architecture of a Long Short-Term Memory (LSTM) network and how it solves vanishing gradients.

Sign in to bookmark

Gated Architectures (LSTM & GRU)

Q4. Compare LSTMs and Gated Recurrent Units (GRUs). What are the trade-offs?

Sign in to bookmark

Sequence-to-Sequence & Attention Mechanisms

Q5. Describe the Encoder-Decoder (Seq2Seq) architecture and its bottleneck.

Sign in to bookmark

Sequence-to-Sequence & Attention Mechanisms

Q6. How does Bahdanau (Additive) Attention solve the Seq2Seq bottleneck?

Sign in to bookmark

Transformers & Self-Attention

Q7. What is Scaled Dot-Product Attention in Transformers, and why is scaling factor $\sqrt{d_k}$ necessary?

Sign in to bookmark

Transformers & Self-Attention

Q8. Why do Transformers require Positional Encodings?

Sign in to bookmark

Training Dynamics & Challenges

Q9. What is Teacher Forcing in sequence training, and what is Exposure Bias?

Sign in to bookmark