Explain Overfitting and Underfitting in Machine Learning
Company: Amazon
Role: Data Scientist
Category: Machine Learning
Difficulty: medium
Interview Round: Technical Screen
##### Scenario
Interview on classical machine-learning fundamentals and computer-vision–related techniques.
##### Question
Differentiate overfitting and underfitting. How do you detect and mitigate each? What is data augmentation? Provide image-specific examples. Describe the main components and purposes of a Convolutional Neural Network. How does an RNN process sequential data? Detail the roles of positional embeddings, self-attention, residual connections and feed-forward networks in a transformer encoder. What is dropout and why does it help? Compare bagging and boosting in terms of bias, variance and algorithm behavior.
##### Hints
Define concepts, mention bias-variance trade-off, regularization tricks, architectures, and practical diagnostics.
Quick Answer: This question evaluates understanding of overfitting and underfitting in machine learning, including recognition of their manifestations, the bias–variance tradeoff, and implications for model generalization in classical ML and computer vision contexts.
