Double Descent Phenomenon Quiz

Answers and explanations appear as you take the quiz.

1. Where does test error typically peak in the double descent curve?
   • At the interpolation threshold where the model just fits all training data
   • At the very start of training with few parameters
   • Only after billions of parameters
   • Test error never peaks; it falls monotonically
2. What happens to test error as a model becomes heavily overparameterized?
   • It rises indefinitely due to overfitting
   • It can decrease again, often below the classical minimum
   • It stays exactly flat
   • It becomes random and unpredictable
3. In which setting does the cited implicit-bias analysis characterize a minimum-norm solution?
   • Any neural network trained with gradient descent
   • Any model that reaches zero training error
   • A single-hidden-layer linear network under stated gradient-flow and initialization assumptions
   • A model trained without an optimization objective
4. 'Epoch-wise' double descent refers to the effect appearing as a function of what?
   • The number of GPUs used
   • The learning-rate schedule only
   • The size of the test set
   • Training time (number of epochs)
5. Which classical idea does double descent challenge?
   • The bias-variance U-shaped tradeoff
   • Backpropagation
   • Stochastic gradient descent
   • The chain rule
6. Who helped popularize and formalize the double descent curve in 2019?
   • Newton and Leibniz
   • Belkin, Hsu, Ma, and Mandal
   • Turing and Church
   • Only anonymous Kaggle competitors
7. Which model families did OpenAI report studying in its deep double-descent experiments?
   • Only decision trees and linear regression
   • Only recurrent speech recognizers
   • Convolutional neural networks, ResNets, and transformers
   • Only graphical game generators
8. Besides model size and training duration, which axis can also show double-descent behavior?
   • Font size in a report
   • The number of words in a product name
   • The screen resolution of a monitor
   • Training dataset size