Data & Machine Learning
Deep Learning
9 practice questions. Free questions open a full answer guide; the rest unlock with Pro.
- What is dropout in a neural network, and why does it help reduce overfitting?
- What is the vanishing gradient problem in deep neural networks, and how is it commonly addressed today?Go Pro
- What does batch normalization do in a neural network, and why does it help training?Go Pro
- You're training a deep network and the training loss either plateaus immediately or blows up to NaN within the first few hundred steps. How do you diagnose which failure you have and systematically work back to a stable run?Go Pro
- Your neural network reaches near-perfect training accuracy but performs poorly on validation data. How would you diagnose and address the overfitting?Go Pro
- What is the difference between batch normalization and layer normalization, and why do transformer architectures use layer normalization?Go Pro
- Your team is training a network that's quite deep, and the early layers barely seem to learn while training stalls. Walk me through what's likely happening and how modern architectures avoid it.Go Pro
- Why do transformer architectures normalize with LayerNorm (or RMSNorm) rather than BatchNorm, and what changes when you place the normalization before the sub-layer (pre-norm) instead of after it (post-norm)?Go Pro
- Walk me through how a sparse mixture-of-experts model works, and the main trade-offs you take on versus a dense model of the same quality.Go Pro
Want questions matched to your role? Paste a job title, job description, or CV for a personalized set, or go Pro to unlock the full bank.