Data for machine learning · Question 55 of 60
A team deliberately adds adversarial examples to the training data of an image classifier. How is this to be classified?
1 answer is correct. Hard
Show answer
Answer
- D As data augmentation in preprocessing to enhance robustness against attacks
Explanation
Preprocessing also includes enriching the data (augmentation): producing more examples, deliberately mixing in adversarial examples so that the model becomes more robust against attacks, and adding synthetic data. This is an intended step, not a mistake.