Continual Learning and Catastrophic Forgetting Quiz

Answers and explanations appear as you take the quiz.

1. What is 'catastrophic forgetting'?
   • When learning a new task sharply degrades performance on previously learned tasks
   • When a model runs out of memory and crashes
   • When training data is accidentally deleted
   • When a model refuses to make predictions
2. The trade-off at the heart of continual learning is often called the:
   • Bias-variance tradeoff
   • Stability-plasticity dilemma
   • Speed-accuracy tradeoff
   • Exploration-exploitation tradeoff
3. What does Elastic Weight Consolidation (EWC) do?
   • Deletes unimportant neurons
   • Doubles the learning rate for new tasks
   • Penalizes changes to weights important for previously learned tasks
   • Stores every past training example forever
4. How do replay-based methods combat forgetting?
   • By freezing the entire network permanently
   • By using a larger learning rate
   • By removing the loss function
   • By interleaving stored or generated old examples into new training
5. How does Gradient Episodic Memory (GEM) constrain a gradient update while learning a new task?
   • It projects the gradient so its first-order approximation does not increase loss on stored prior-task examples
   • It replaces all prior-task data with new examples
   • It freezes every parameter for all future tasks
   • It increases weights that were unimportant to earlier tasks
6. Why does naive fine-tuning on a new task cause forgetting?
   • The GPU overheats and corrupts memory
   • Gradient descent moves shared weights toward the new task with no constraint to preserve old ones
   • New tasks always have more data
   • The model intentionally deletes old weights
7. How does iCaRL classify a test example after learning multiple classes?
   • By retrieving the single closest word in a text corpus
   • By selecting the class with the largest Fisher information
   • By comparing its feature vector with per-class means computed from stored exemplars
   • By replaying every training image at inference
8. Why is evaluating continual learning more than measuring final accuracy on the newest task?
   • New-task accuracy is the only relevant measure
   • Old tasks are removed from the evaluation
   • The model cannot be tested on multiple tasks
   • A method must also measure retention on prior tasks and its ability to learn new ones